Merge

47 files changed, 751 insertions, 387 deletions

diff --git a/src/share/vm/adlc/formssel.cpp b/src/share/vm/adlc/formssel.cpp
index 27651dc68..3d4a562e4 100644
--- a/src/share/vm/adlc/formssel.cpp
+++ b/src/share/vm/adlc/formssel.cpp
@@ -235,6 +235,9 @@ bool InstructForm::is_parm(FormDict &globals) {
   return false;
 }
 
+bool InstructForm::is_ideal_negD() const {
+  return (_matrule && _matrule->_rChild && strcmp(_matrule->_rChild->_opType, "NegD") == 0);
+}
 
 // Return 'true' if this instruction matches an ideal 'Copy*' node
 int InstructForm::is_ideal_copy() const {
@@ -533,6 +536,12 @@ bool InstructForm::rematerialize(FormDict &globals, RegisterForm *registers ) {
   if( data_type != Form::none )
     rematerialize = true;
 
+  // Ugly: until a better fix is implemented, disable rematerialization for
+  // negD nodes because they are proved to be problematic.
+  if (is_ideal_negD()) {
+    return false;
+  }
+
   // Constants
   if( _components.count() == 1 && _components[0]->is(Component::USE_DEF) )
     rematerialize = true;
diff --git a/src/share/vm/adlc/formssel.hpp b/src/share/vm/adlc/formssel.hpp
index 622d8902f..fa3e9ff8f 100644
--- a/src/share/vm/adlc/formssel.hpp
+++ b/src/share/vm/adlc/formssel.hpp
@@ -147,6 +147,7 @@ public:
   virtual int         is_empty_encoding() const; // _size=0 and/or _insencode empty
   virtual int         is_tls_instruction() const; // tlsLoadP rule or ideal ThreadLocal
   virtual int         is_ideal_copy() const;    // node matches ideal 'Copy*'
+  virtual bool        is_ideal_negD() const;    // node matches ideal 'NegD'
   virtual bool        is_ideal_if()   const;    // node matches ideal 'If'
   virtual bool        is_ideal_fastlock() const; // node matches 'FastLock'
   virtual bool        is_ideal_membar() const;  // node matches ideal 'MemBarXXX'
diff --git a/src/share/vm/c1/c1_IR.cpp b/src/share/vm/c1/c1_IR.cpp
index e9e73db0c..13a7f790f 100644
--- a/src/share/vm/c1/c1_IR.cpp
+++ b/src/share/vm/c1/c1_IR.cpp
@@ -506,7 +506,7 @@ ComputeLinearScanOrder::ComputeLinearScanOrder(Compilation* c, BlockBegin* start
   _loop_map(0, 0),          // initialized later with correct size
   _compilation(c)
 {
-  TRACE_LINEAR_SCAN(2, "***** computing linear-scan block order");
+  TRACE_LINEAR_SCAN(2, tty->print_cr("***** computing linear-scan block order"));
 
   init_visited();
   count_edges(start_block, NULL);
@@ -683,7 +683,7 @@ void ComputeLinearScanOrder::clear_non_natural_loops(BlockBegin* start_block) {
 }
 
 void ComputeLinearScanOrder::assign_loop_depth(BlockBegin* start_block) {
-  TRACE_LINEAR_SCAN(3, "----- computing loop-depth and weight");
+  TRACE_LINEAR_SCAN(3, tty->print_cr("----- computing loop-depth and weight"));
   init_visited();
 
   assert(_work_list.is_empty(), "work list must be empty before processing");
@@ -868,7 +868,7 @@ void ComputeLinearScanOrder::append_block(BlockBegin* cur) {
 }
 
 void ComputeLinearScanOrder::compute_order(BlockBegin* start_block) {
-  TRACE_LINEAR_SCAN(3, "----- computing final block order");
+  TRACE_LINEAR_SCAN(3, tty->print_cr("----- computing final block order"));
 
   // the start block is always the first block in the linear scan order
   _linear_scan_order = new BlockList(_num_blocks);
diff --git a/src/share/vm/c1/c1_LIR.cpp b/src/share/vm/c1/c1_LIR.cpp
index f26d1812c..e2611534e 100644
--- a/src/share/vm/c1/c1_LIR.cpp
+++ b/src/share/vm/c1/c1_LIR.cpp
@@ -201,23 +201,24 @@ void LIR_OprDesc::validate_type() const {
 
 #ifdef ASSERT
   if (!is_pointer() && !is_illegal()) {
+    OprKind kindfield = kind_field(); // Factored out because of compiler bug, see 8002160
     switch (as_BasicType(type_field())) {
     case T_LONG:
-      assert((kind_field() == cpu_register || kind_field() == stack_value) &&
+      assert((kindfield == cpu_register || kindfield == stack_value) &&
              size_field() == double_size, "must match");
       break;
     case T_FLOAT:
       // FP return values can be also in CPU registers on ARM and PPC (softfp ABI)
-      assert((kind_field() == fpu_register || kind_field() == stack_value
-             ARM_ONLY(|| kind_field() == cpu_register)
-             PPC_ONLY(|| kind_field() == cpu_register) ) &&
+      assert((kindfield == fpu_register || kindfield == stack_value
+             ARM_ONLY(|| kindfield == cpu_register)
+             PPC_ONLY(|| kindfield == cpu_register) ) &&
              size_field() == single_size, "must match");
       break;
     case T_DOUBLE:
       // FP return values can be also in CPU registers on ARM and PPC (softfp ABI)
-      assert((kind_field() == fpu_register || kind_field() == stack_value
-             ARM_ONLY(|| kind_field() == cpu_register)
-             PPC_ONLY(|| kind_field() == cpu_register) ) &&
+      assert((kindfield == fpu_register || kindfield == stack_value
+             ARM_ONLY(|| kindfield == cpu_register)
+             PPC_ONLY(|| kindfield == cpu_register) ) &&
              size_field() == double_size, "must match");
       break;
     case T_BOOLEAN:
@@ -229,7 +230,7 @@ void LIR_OprDesc::validate_type() const {
     case T_OBJECT:
     case T_METADATA:
     case T_ARRAY:
-      assert((kind_field() == cpu_register || kind_field() == stack_value) &&
+      assert((kindfield == cpu_register || kindfield == stack_value) &&
              size_field() == single_size, "must match");
       break;
 
diff --git a/src/share/vm/ci/ciUtilities.hpp b/src/share/vm/ci/ciUtilities.hpp
index 1a075bf6e..2032a8f35 100644
--- a/src/share/vm/ci/ciUtilities.hpp
+++ b/src/share/vm/ci/ciUtilities.hpp
@@ -96,7 +96,7 @@
     CLEAR_PENDING_EXCEPTION;                     \
     return (result);                             \
   }                                              \
-  (0
+  (void)(0
 
 #define KILL_COMPILE_ON_ANY                      \
   THREAD);                                       \
@@ -104,7 +104,7 @@
     fatal("unhandled ci exception");             \
     CLEAR_PENDING_EXCEPTION;                     \
   }                                              \
-(0
+(void)(0
 
 
 inline const char* bool_to_str(bool b) {
diff --git a/src/share/vm/classfile/genericSignatures.cpp b/src/share/vm/classfile/genericSignatures.cpp
index 33fbca051..3bbce2f8c 100644
--- a/src/share/vm/classfile/genericSignatures.cpp
+++ b/src/share/vm/classfile/genericSignatures.cpp
@@ -124,7 +124,7 @@ class DescriptorStream : public ResourceObj {
       fatal(STREAM->parse_error());      \
     }                                   \
     return NULL;                        \
-  } 0
+  } (void)0
 
 #define READ() STREAM->read(); CHECK_FOR_PARSE_ERROR()
 #define PEEK() STREAM->peek(); CHECK_FOR_PARSE_ERROR()
@@ -133,7 +133,7 @@ class DescriptorStream : public ResourceObj {
 #define EXPECTED(c, ch) STREAM->assert_char(c, ch); CHECK_FOR_PARSE_ERROR()
 #define EXPECT_END() STREAM->expect_end(); CHECK_FOR_PARSE_ERROR()
 
-#define CHECK_STREAM STREAM); CHECK_FOR_PARSE_ERROR(); (0
+#define CHECK_STREAM STREAM); CHECK_FOR_PARSE_ERROR(); ((void)0
 
 #ifndef PRODUCT
 void Identifier::print_on(outputStream* str) const {
diff --git a/src/share/vm/classfile/symbolTable.cpp b/src/share/vm/classfile/symbolTable.cpp
index c78ea2d35..c00e95817 100644
--- a/src/share/vm/classfile/symbolTable.cpp
+++ b/src/share/vm/classfile/symbolTable.cpp
@@ -598,6 +598,8 @@ StringTable* StringTable::_the_table = NULL;
 
 bool StringTable::_needs_rehashing = false;
 
+volatile int StringTable::_parallel_claimed_idx = 0;
+
 // Pick hashing algorithm
 unsigned int StringTable::hash_string(const jchar* s, int len) {
   return use_alternate_hashcode() ? AltHashing::murmur3_32(seed(), s, len) :
@@ -761,8 +763,18 @@ void StringTable::unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f)
   }
 }
 
-void StringTable::oops_do(OopClosure* f) {
-  for (int i = 0; i < the_table()->table_size(); ++i) {
+void StringTable::buckets_do(OopClosure* f, int start_idx, int end_idx) {
+  const int limit = the_table()->table_size();
+
+  assert(0 <= start_idx && start_idx <= limit,
+         err_msg("start_idx (" INT32_FORMAT ") oob?", start_idx));
+  assert(0 <= end_idx && end_idx <= limit,
+         err_msg("end_idx (" INT32_FORMAT ") oob?", end_idx));
+  assert(start_idx <= end_idx,
+         err_msg("Ordering: start_idx=" INT32_FORMAT", end_idx=" INT32_FORMAT,
+                 start_idx, end_idx));
+
+  for (int i = start_idx; i < end_idx; i += 1) {
     HashtableEntry<oop, mtSymbol>* entry = the_table()->bucket(i);
     while (entry != NULL) {
       assert(!entry->is_shared(), "CDS not used for the StringTable");
@@ -774,6 +786,27 @@ void StringTable::oops_do(OopClosure* f) {
   }
 }
 
+void StringTable::oops_do(OopClosure* f) {
+  buckets_do(f, 0, the_table()->table_size());
+}
+
+void StringTable::possibly_parallel_oops_do(OopClosure* f) {
+  const int ClaimChunkSize = 32;
+  const int limit = the_table()->table_size();
+
+  for (;;) {
+    // Grab next set of buckets to scan
+    int start_idx = Atomic::add(ClaimChunkSize, &_parallel_claimed_idx) - ClaimChunkSize;
+    if (start_idx >= limit) {
+      // End of table
+      break;
+    }
+
+    int end_idx = MIN2(limit, start_idx + ClaimChunkSize);
+    buckets_do(f, start_idx, end_idx);
+  }
+}
+
 void StringTable::verify() {
   for (int i = 0; i < the_table()->table_size(); ++i) {
     HashtableEntry<oop, mtSymbol>* p = the_table()->bucket(i);
diff --git a/src/share/vm/classfile/symbolTable.hpp b/src/share/vm/classfile/symbolTable.hpp
index 4701e74b5..7a0031c9a 100644
--- a/src/share/vm/classfile/symbolTable.hpp
+++ b/src/share/vm/classfile/symbolTable.hpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -246,12 +246,19 @@ private:
   // Set if one bucket is out of balance due to hash algorithm deficiency
   static bool _needs_rehashing;
 
+  // Claimed high water mark for parallel chunked scanning
+  static volatile int _parallel_claimed_idx;
+
   static oop intern(Handle string_or_null, jchar* chars, int length, TRAPS);
   oop basic_add(int index, Handle string_or_null, jchar* name, int len,
                 unsigned int hashValue, TRAPS);
 
   oop lookup(int index, jchar* chars, int length, unsigned int hashValue);
 
+  // Apply the give oop closure to the entries to the buckets
+  // in the range [start_idx, end_idx).
+  static void buckets_do(OopClosure* f, int start_idx, int end_idx);
+
   StringTable() : Hashtable<oop, mtSymbol>((int)StringTableSize,
                               sizeof (HashtableEntry<oop, mtSymbol>)) {}
 
@@ -277,9 +284,12 @@ public:
     unlink_or_oops_do(cl, NULL);
   }
 
-  // Invoke "f->do_oop" on the locations of all oops in the table.
+  // Serially invoke "f->do_oop" on the locations of all oops in the table.
   static void oops_do(OopClosure* f);
 
+  // Possibly parallel version of the above
+  static void possibly_parallel_oops_do(OopClosure* f);
+
   // Hashing algorithm, used as the hash value used by the
   //     StringTable for bucket selection and comparison (stored in the
   //     HashtableEntry structures).  This is used in the String.intern() method.
@@ -315,5 +325,8 @@ public:
   // Rehash the symbol table if it gets out of balance
   static void rehash_table();
   static bool needs_rehashing() { return _needs_rehashing; }
+
+  // Parallel chunked scanning
+  static void clear_parallel_claimed_index() { _parallel_claimed_idx = 0; }
 };
 #endif // SHARE_VM_CLASSFILE_SYMBOLTABLE_HPP
diff --git a/src/share/vm/classfile/verifier.hpp b/src/share/vm/classfile/verifier.hpp
index 34a1e0b3a..9da0ac34a 100644
--- a/src/share/vm/classfile/verifier.hpp
+++ b/src/share/vm/classfile/verifier.hpp
@@ -86,9 +86,9 @@ class StackMapTable;
 // These macros are used similarly to CHECK macros but also check
 // the status of the verifier and return if that has an error.
 #define CHECK_VERIFY(verifier) \
-  CHECK); if ((verifier)->has_error()) return; (0
+  CHECK); if ((verifier)->has_error()) return; ((void)0
 #define CHECK_VERIFY_(verifier, result) \
-  CHECK_(result)); if ((verifier)->has_error()) return (result); (0
+  CHECK_(result)); if ((verifier)->has_error()) return (result); ((void)0
 
 class TypeOrigin VALUE_OBJ_CLASS_SPEC {
  private:
diff --git a/src/share/vm/code/dependencies.cpp b/src/share/vm/code/dependencies.cpp
index b790c3597..4db37819e 100644
--- a/src/share/vm/code/dependencies.cpp
+++ b/src/share/vm/code/dependencies.cpp
@@ -989,7 +989,7 @@ Klass* ClassHierarchyWalker::find_witness_in(KlassDepChange& changes,
   assert(changes.involves_context(context_type), "irrelevant dependency");
   Klass* new_type = changes.new_type();
 
-  count_find_witness_calls();
+  (void)count_find_witness_calls();
   NOT_PRODUCT(deps_find_witness_singles++);
 
   // Current thread must be in VM (not native mode, as in CI):
diff --git a/src/share/vm/code/nmethod.cpp b/src/share/vm/code/nmethod.cpp
index 53f8cd07b..41eb628fb 100644
--- a/src/share/vm/code/nmethod.cpp
+++ b/src/share/vm/code/nmethod.cpp
@@ -2615,7 +2615,8 @@ void nmethod::print_relocations() {
                       relocation_begin()-1+ip[1]);
       for (; ip < index_end; ip++)
         tty->print_cr("  (%d ?)", ip[0]);
-      tty->print_cr("          @" INTPTR_FORMAT ": index_size=%d", ip, *ip++);
+      tty->print_cr("          @" INTPTR_FORMAT ": index_size=%d", ip, *ip);
+      ip++;
       tty->print_cr("reloc_end @" INTPTR_FORMAT ":", ip);
     }
   }
diff --git a/src/share/vm/gc_implementation/parallelScavenge/cardTableExtension.cpp b/src/share/vm/gc_implementation/parallelScavenge/cardTableExtension.cpp
index b2134f10e..82ea39b52 100644
--- a/src/share/vm/gc_implementation/parallelScavenge/cardTableExtension.cpp
+++ b/src/share/vm/gc_implementation/parallelScavenge/cardTableExtension.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -565,11 +565,9 @@ bool CardTableExtension::resize_commit_uncommit(int changed_region,
     if(new_start_aligned < new_end_for_commit) {
       MemRegion new_committed =
         MemRegion(new_start_aligned, new_end_for_commit);
-      if (!os::commit_memory((char*)new_committed.start(),
-                             new_committed.byte_size())) {
-        vm_exit_out_of_memory(new_committed.byte_size(), OOM_MMAP_ERROR,
-                              "card table expansion");
-      }
+      os::commit_memory_or_exit((char*)new_committed.start(),
+                                new_committed.byte_size(), !ExecMem,
+                                "card table expansion");
     }
     result = true;
   } else if (new_start_aligned > cur_committed.start()) {
diff --git a/src/share/vm/gc_implementation/parallelScavenge/psAdaptiveSizePolicy.cpp b/src/share/vm/gc_implementation/parallelScavenge/psAdaptiveSizePolicy.cpp
index 772701795..7fba4f358 100644
--- a/src/share/vm/gc_implementation/parallelScavenge/psAdaptiveSizePolicy.cpp
+++ b/src/share/vm/gc_implementation/parallelScavenge/psAdaptiveSizePolicy.cpp
@@ -1250,14 +1250,13 @@ uint PSAdaptiveSizePolicy::compute_survivor_space_size_and_threshold(
                   avg_promoted()->deviation());
     }
 
-    gclog_or_tty->print( "  avg_promoted_padded_avg: %f"
+    gclog_or_tty->print_cr( "  avg_promoted_padded_avg: %f"
                 "  avg_pretenured_padded_avg: %f"
                 "  tenuring_thresh: %d"
                 "  target_size: " SIZE_FORMAT,
                 avg_promoted()->padded_average(),
                 _avg_pretenured->padded_average(),
                 tenuring_threshold, target_size);
-    tty->cr();
   }
 
   set_survivor_size(target_size);
@@ -1279,7 +1278,7 @@ void PSAdaptiveSizePolicy::update_averages(bool is_survivor_overflow,
   avg_promoted()->sample(promoted + _avg_pretenured->padded_average());
 
   if (PrintAdaptiveSizePolicy) {
-    gclog_or_tty->print(
+    gclog_or_tty->print_cr(
                   "AdaptiveSizePolicy::update_averages:"
                   "  survived: "  SIZE_FORMAT
                   "  promoted: "  SIZE_FORMAT
diff --git a/src/share/vm/gc_implementation/parallelScavenge/psVirtualspace.cpp b/src/share/vm/gc_implementation/parallelScavenge/psVirtualspace.cpp
index e19476707..999952730 100644
--- a/src/share/vm/gc_implementation/parallelScavenge/psVirtualspace.cpp
+++ b/src/share/vm/gc_implementation/parallelScavenge/psVirtualspace.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -101,7 +101,8 @@ bool PSVirtualSpace::expand_by(size_t bytes) {
   }
 
   char* const base_addr = committed_high_addr();
-  bool result = special() || os::commit_memory(base_addr, bytes, alignment());
+  bool result = special() ||
+         os::commit_memory(base_addr, bytes, alignment(), !ExecMem);
   if (result) {
     _committed_high_addr += bytes;
   }
@@ -154,7 +155,7 @@ PSVirtualSpace::expand_into(PSVirtualSpace* other_space, size_t bytes) {
   if (tmp_bytes > 0) {
     char* const commit_base = committed_high_addr();
     if (other_space->special() ||
-        os::commit_memory(commit_base, tmp_bytes, alignment())) {
+        os::commit_memory(commit_base, tmp_bytes, alignment(), !ExecMem)) {
       // Reduce the reserved region in the other space.
       other_space->set_reserved(other_space->reserved_low_addr() + tmp_bytes,
                                 other_space->reserved_high_addr(),
@@ -269,7 +270,8 @@ bool PSVirtualSpaceHighToLow::expand_by(size_t bytes) {
   }
 
   char* const base_addr = committed_low_addr() - bytes;
-  bool result = special() || os::commit_memory(base_addr, bytes, alignment());
+  bool result = special() ||
+         os::commit_memory(base_addr, bytes, alignment(), !ExecMem);
   if (result) {
     _committed_low_addr -= bytes;
   }
@@ -322,7 +324,7 @@ size_t PSVirtualSpaceHighToLow::expand_into(PSVirtualSpace* other_space,
   if (tmp_bytes > 0) {
     char* const commit_base = committed_low_addr() - tmp_bytes;
     if (other_space->special() ||
-        os::commit_memory(commit_base, tmp_bytes, alignment())) {
+        os::commit_memory(commit_base, tmp_bytes, alignment(), !ExecMem)) {
       // Reduce the reserved region in the other space.
       other_space->set_reserved(other_space->reserved_low_addr(),
                                 other_space->reserved_high_addr() - tmp_bytes,
diff --git a/src/share/vm/memory/allocation.hpp b/src/share/vm/memory/allocation.hpp
index bd38a94d4..f56625c7f 100644
--- a/src/share/vm/memory/allocation.hpp
+++ b/src/share/vm/memory/allocation.hpp
@@ -643,8 +643,15 @@ class ResourceObj ALLOCATION_SUPER_CLASS_SPEC {
 #define NEW_RESOURCE_ARRAY_IN_THREAD(thread, type, size)\
   (type*) resource_allocate_bytes(thread, (size) * sizeof(type))
 
+#define NEW_RESOURCE_ARRAY_IN_THREAD_RETURN_NULL(thread, type, size)\
+  (type*) resource_allocate_bytes(thread, (size) * sizeof(type), AllocFailStrategy::RETURN_NULL)
+
 #define REALLOC_RESOURCE_ARRAY(type, old, old_size, new_size)\
-  (type*) resource_reallocate_bytes((char*)(old), (old_size) * sizeof(type), (new_size) * sizeof(type) )
+  (type*) resource_reallocate_bytes((char*)(old), (old_size) * sizeof(type), (new_size) * sizeof(type))
+
+#define REALLOC_RESOURCE_ARRAY_RETURN_NULL(type, old, old_size, new_size)\
+  (type*) resource_reallocate_bytes((char*)(old), (old_size) * sizeof(type),\
+                                    (new_size) * sizeof(type), AllocFailStrategy::RETURN_NULL)
 
 #define FREE_RESOURCE_ARRAY(type, old, size)\
   resource_free_bytes((char*)(old), (size) * sizeof(type))
@@ -655,28 +662,40 @@ class ResourceObj ALLOCATION_SUPER_CLASS_SPEC {
 #define NEW_RESOURCE_OBJ(type)\
   NEW_RESOURCE_ARRAY(type, 1)
 
+#define NEW_RESOURCE_OBJ_RETURN_NULL(type)\
+  NEW_RESOURCE_ARRAY_RETURN_NULL(type, 1)
+
+#define NEW_C_HEAP_ARRAY3(type, size, memflags, pc, allocfail)\
+  (type*) AllocateHeap(size * sizeof(type), memflags, pc, allocfail)
+
+#define NEW_C_HEAP_ARRAY2(type, size, memflags, pc)\
+  (type*) (AllocateHeap((size) * sizeof(type), memflags, pc))
+
 #define NEW_C_HEAP_ARRAY(type, size, memflags)\
   (type*) (AllocateHeap((size) * sizeof(type), memflags))
 
+#define NEW_C_HEAP_ARRAY2_RETURN_NULL(type, size, memflags, pc)\
+  NEW_C_HEAP_ARRAY3(type, size, memflags, pc, AllocFailStrategy::RETURN_NULL)
+
+#define NEW_C_HEAP_ARRAY_RETURN_NULL(type, size, memflags)\
+  NEW_C_HEAP_ARRAY3(type, size, memflags, (address)0, AllocFailStrategy::RETURN_NULL)
+
 #define REALLOC_C_HEAP_ARRAY(type, old, size, memflags)\
   (type*) (ReallocateHeap((char*)old, (size) * sizeof(type), memflags))
 
+#define REALLOC_C_HEAP_ARRAY_RETURN_NULL(type, old, size, memflags)\
+   (type*) (ReallocateHeap((char*)old, (size) * sizeof(type), memflags, AllocFailStrategy::RETURN_NULL))
+
 #define FREE_C_HEAP_ARRAY(type, old, memflags) \
   FreeHeap((char*)(old), memflags)
 
-#define NEW_C_HEAP_ARRAY2(type, size, memflags, pc)\
-  (type*) (AllocateHeap((size) * sizeof(type), memflags, pc))
-
-#define REALLOC_C_HEAP_ARRAY2(type, old, size, memflags, pc)\
-  (type*) (ReallocateHeap((char*)old, (size) * sizeof(type), memflags, pc))
-
-#define NEW_C_HEAP_ARRAY3(type, size, memflags, pc, allocfail)         \
-  (type*) AllocateHeap(size * sizeof(type), memflags, pc, allocfail)
-
 // allocate type in heap without calling ctor
 #define NEW_C_HEAP_OBJ(type, memflags)\
   NEW_C_HEAP_ARRAY(type, 1, memflags)
 
+#define NEW_C_HEAP_OBJ_RETURN_NULL(type, memflags)\
+  NEW_C_HEAP_ARRAY_RETURN_NULL(type, 1, memflags)
+
 // deallocate obj of type in heap without calling dtor
 #define FREE_C_HEAP_OBJ(objname, memflags)\
   FreeHeap((char*)objname, memflags);
@@ -721,13 +740,21 @@ public:
 // is set so that we always use malloc except for Solaris where we set the
 // limit to get mapped memory.
 template <class E, MEMFLAGS F>
-class ArrayAllocator : StackObj {
+class ArrayAllocator VALUE_OBJ_CLASS_SPEC {
   char* _addr;
   bool _use_malloc;
   size_t _size;
+  bool _free_in_destructor;
  public:
-  ArrayAllocator() : _addr(NULL), _use_malloc(false), _size(0) { }
-  ~ArrayAllocator() { free(); }
+  ArrayAllocator(bool free_in_destructor = true) :
+    _addr(NULL), _use_malloc(false), _size(0), _free_in_destructor(free_in_destructor) { }
+
+  ~ArrayAllocator() {
+    if (_free_in_destructor) {
+      free();
+    }
+  }
+
   E* allocate(size_t length);
   void free();
 };
diff --git a/src/share/vm/memory/allocation.inline.hpp b/src/share/vm/memory/allocation.inline.hpp
index 138dd660a..f59ae5b4f 100644
--- a/src/share/vm/memory/allocation.inline.hpp
+++ b/src/share/vm/memory/allocation.inline.hpp
@@ -146,10 +146,7 @@ E* ArrayAllocator<E, F>::allocate(size_t length) {
     vm_exit_out_of_memory(_size, OOM_MMAP_ERROR, "Allocator (reserve)");
   }
 
-  bool success = os::commit_memory(_addr, _size, false /* executable */);
-  if (!success) {
-    vm_exit_out_of_memory(_size, OOM_MMAP_ERROR, "Allocator (commit)");
-  }
+  os::commit_memory_or_exit(_addr, _size, !ExecMem, "Allocator (commit)");
 
   return (E*)_addr;
 }
diff --git a/src/share/vm/memory/cardTableModRefBS.cpp b/src/share/vm/memory/cardTableModRefBS.cpp
index 7cc68debf..82336503e 100644
--- a/src/share/vm/memory/cardTableModRefBS.cpp
+++ b/src/share/vm/memory/cardTableModRefBS.cpp
@@ -110,11 +110,8 @@ CardTableModRefBS::CardTableModRefBS(MemRegion whole_heap,
   jbyte* guard_card = &_byte_map[_guard_index];
   uintptr_t guard_page = align_size_down((uintptr_t)guard_card, _page_size);
   _guard_region = MemRegion((HeapWord*)guard_page, _page_size);
-  if (!os::commit_memory((char*)guard_page, _page_size, _page_size)) {
-    // Do better than this for Merlin
-    vm_exit_out_of_memory(_page_size, OOM_MMAP_ERROR, "card table last card");
-  }
-
+  os::commit_memory_or_exit((char*)guard_page, _page_size, _page_size,
+                            !ExecMem, "card table last card");
   *guard_card = last_card;
 
    _lowest_non_clean =
@@ -312,12 +309,9 @@ void CardTableModRefBS::resize_covered_region(MemRegion new_region) {
         MemRegion(cur_committed.end(), new_end_for_commit);
 
       assert(!new_committed.is_empty(), "Region should not be empty here");
-      if (!os::commit_memory((char*)new_committed.start(),
-                             new_committed.byte_size(), _page_size)) {
-        // Do better than this for Merlin
-        vm_exit_out_of_memory(new_committed.byte_size(), OOM_MMAP_ERROR,
-                "card table expansion");
-      }
+      os::commit_memory_or_exit((char*)new_committed.start(),
+                                new_committed.byte_size(), _page_size,
+                                !ExecMem, "card table expansion");
     // Use new_end_aligned (as opposed to new_end_for_commit) because
     // the cur_committed region may include the guard region.
     } else if (new_end_aligned < cur_committed.end()) {
@@ -418,7 +412,7 @@ void CardTableModRefBS::resize_covered_region(MemRegion new_region) {
   }
   // Touch the last card of the covered region to show that it
   // is committed (or SEGV).
-  debug_only(*byte_for(_covered[ind].last());)
+  debug_only((void) (*byte_for(_covered[ind].last()));)
   debug_only(verify_guard();)
 }
 
diff --git a/src/share/vm/memory/sharedHeap.cpp b/src/share/vm/memory/sharedHeap.cpp
index 1dfccd0b5..fa4dd64ec 100644
--- a/src/share/vm/memory/sharedHeap.cpp
+++ b/src/share/vm/memory/sharedHeap.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -47,7 +47,6 @@ enum SH_process_strong_roots_tasks {
   SH_PS_SystemDictionary_oops_do,
   SH_PS_ClassLoaderDataGraph_oops_do,
   SH_PS_jvmti_oops_do,
-  SH_PS_StringTable_oops_do,
   SH_PS_CodeCache_oops_do,
   // Leave this one last.
   SH_PS_NumElements
@@ -127,6 +126,8 @@ SharedHeap::StrongRootsScope::StrongRootsScope(SharedHeap* outer, bool activate)
 {
   if (_active) {
     outer->change_strong_roots_parity();
+    // Zero the claimed high water mark in the StringTable
+    StringTable::clear_parallel_claimed_index();
   }
 }
 
@@ -154,14 +155,16 @@ void SharedHeap::process_strong_roots(bool activate_scope,
   // Global (strong) JNI handles
   if (!_process_strong_tasks->is_task_claimed(SH_PS_JNIHandles_oops_do))
     JNIHandles::oops_do(roots);
+
   // All threads execute this; the individual threads are task groups.
   CLDToOopClosure roots_from_clds(roots);
   CLDToOopClosure* roots_from_clds_p = (is_scavenging ? NULL : &roots_from_clds);
-  if (ParallelGCThreads > 0) {
-    Threads::possibly_parallel_oops_do(roots, roots_from_clds_p ,code_roots);
+  if (CollectedHeap::use_parallel_gc_threads()) {
+    Threads::possibly_parallel_oops_do(roots, roots_from_clds_p, code_roots);
   } else {
     Threads::oops_do(roots, roots_from_clds_p, code_roots);
   }
+
   if (!_process_strong_tasks-> is_task_claimed(SH_PS_ObjectSynchronizer_oops_do))
     ObjectSynchronizer::oops_do(roots);
   if (!_process_strong_tasks->is_task_claimed(SH_PS_FlatProfiler_oops_do))
@@ -189,8 +192,12 @@ void SharedHeap::process_strong_roots(bool activate_scope,
     }
   }
 
-  if (!_process_strong_tasks->is_task_claimed(SH_PS_StringTable_oops_do)) {
-    if (so & SO_Strings) {
+  // All threads execute the following. A specific chunk of buckets
+  // from the StringTable are the individual tasks.
+  if (so & SO_Strings) {
+    if (CollectedHeap::use_parallel_gc_threads()) {
+      StringTable::possibly_parallel_oops_do(roots);
+    } else {
       StringTable::oops_do(roots);
     }
   }
diff --git a/src/share/vm/memory/universe.cpp b/src/share/vm/memory/universe.cpp
index 2b4c5602a..fbd1cdafc 100644
--- a/src/share/vm/memory/universe.cpp
+++ b/src/share/vm/memory/universe.cpp
@@ -531,7 +531,9 @@ void Universe::reinitialize_vtable_of(KlassHandle k_h, TRAPS) {
   if (vt) vt->initialize_vtable(false, CHECK);
   if (ko->oop_is_instance()) {
     InstanceKlass* ik = (InstanceKlass*)ko;
-    for (KlassHandle s_h(THREAD, ik->subklass()); s_h() != NULL; s_h = (THREAD, s_h()->next_sibling())) {
+    for (KlassHandle s_h(THREAD, ik->subklass());
+         s_h() != NULL;
+         s_h = KlassHandle(THREAD, s_h()->next_sibling())) {
       reinitialize_vtable_of(s_h, CHECK);
     }
   }
diff --git a/src/share/vm/opto/c2_globals.hpp b/src/share/vm/opto/c2_globals.hpp
index 49ed1e8bc..2e212995f 100644
--- a/src/share/vm/opto/c2_globals.hpp
+++ b/src/share/vm/opto/c2_globals.hpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -406,10 +406,10 @@
   develop(intx, WarmCallMaxSize, 999999,                                    \
           "size of the largest inlinable method")                           \
                                                                             \
-  product(intx, MaxNodeLimit, 65000,                                        \
+  product(intx, MaxNodeLimit, 80000,                                        \
           "Maximum number of nodes")                                        \
                                                                             \
-  product(intx, NodeLimitFudgeFactor, 1000,                                 \
+  product(intx, NodeLimitFudgeFactor, 2000,                                 \
           "Fudge Factor for certain optimizations")                         \
                                                                             \
   product(bool, UseJumpTables, true,                                        \
diff --git a/src/share/vm/opto/chaitin.cpp b/src/share/vm/opto/chaitin.cpp
index 9d69b0f3b..53ffc5732 100644
--- a/src/share/vm/opto/chaitin.cpp
+++ b/src/share/vm/opto/chaitin.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -435,6 +435,9 @@ void PhaseChaitin::Register_Allocate() {
     // Insert un-coalesced copies.  Visit all Phis.  Where inputs to a Phi do
     // not match the Phi itself, insert a copy.
     coalesce.insert_copies(_matcher);
+    if (C->failing()) {
+      return;
+    }
   }
 
   // After aggressive coalesce, attempt a first cut at coloring.
diff --git a/src/share/vm/opto/coalesce.cpp b/src/share/vm/opto/coalesce.cpp
index 74618fb41..60c88dc12 100644
--- a/src/share/vm/opto/coalesce.cpp
+++ b/src/share/vm/opto/coalesce.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -240,6 +240,8 @@ void PhaseAggressiveCoalesce::insert_copies( Matcher &matcher ) {
   _unique = C->unique();
 
   for( uint i=0; i<_phc._cfg._num_blocks; i++ ) {
+    C->check_node_count(NodeLimitFudgeFactor, "out of nodes in coalesce");
+    if (C->failing()) return;
     Block *b = _phc._cfg._blocks[i];
     uint cnt = b->num_preds();  // Number of inputs to the Phi
 
diff --git a/src/share/vm/opto/matcher.cpp b/src/share/vm/opto/matcher.cpp
index 4f3aad763..58de9b625 100644
--- a/src/share/vm/opto/matcher.cpp
+++ b/src/share/vm/opto/matcher.cpp
@@ -985,6 +985,8 @@ Node *Matcher::xform( Node *n, int max_stack ) {
   mstack.push(n, Visit, NULL, -1);  // set NULL as parent to indicate root
 
   while (mstack.is_nonempty()) {
+    C->check_node_count(NodeLimitFudgeFactor, "too many nodes matching instructions");
+    if (C->failing()) return NULL;
     n = mstack.node();          // Leave node on stack
     Node_State nstate = mstack.state();
     if (nstate == Visit) {
diff --git a/src/share/vm/opto/memnode.cpp b/src/share/vm/opto/memnode.cpp
index fc59cebdc..30acf3513 100644
--- a/src/share/vm/opto/memnode.cpp
+++ b/src/share/vm/opto/memnode.cpp
@@ -2930,7 +2930,9 @@ MemBarNode* MemBarNode::make(Compile* C, int opcode, int atp, Node* pn) {
 Node *MemBarNode::Ideal(PhaseGVN *phase, bool can_reshape) {
   if (remove_dead_region(phase, can_reshape)) return this;
   // Don't bother trying to transform a dead node
-  if (in(0) && in(0)->is_top())  return NULL;
+  if (in(0) && in(0)->is_top()) {
+    return NULL;
+  }
 
   // Eliminate volatile MemBars for scalar replaced objects.
   if (can_reshape && req() == (Precedent+1)) {
@@ -2939,6 +2941,14 @@ Node *MemBarNode::Ideal(PhaseGVN *phase, bool can_reshape) {
     if ((opc == Op_MemBarAcquire || opc == Op_MemBarVolatile)) {
       // Volatile field loads and stores.
       Node* my_mem = in(MemBarNode::Precedent);
+      // The MembarAquire may keep an unused LoadNode alive through the Precedent edge
+      if ((my_mem != NULL) && (opc == Op_MemBarAcquire) && (my_mem->outcnt() == 1)) {
+        assert(my_mem->unique_out() == this, "sanity");
+        phase->hash_delete(this);
+        del_req(Precedent);
+        phase->is_IterGVN()->_worklist.push(my_mem); // remove dead node later
+        my_mem = NULL;
+      }
       if (my_mem != NULL && my_mem->is_Mem()) {
         const TypeOopPtr* t_oop = my_mem->in(MemNode::Address)->bottom_type()->isa_oopptr();
         // Check for scalar replaced object reference.
@@ -4384,7 +4394,7 @@ static void verify_memory_slice(const MergeMemNode* m, int alias_idx, Node* n) {
   }
 }
 #else // !ASSERT
-#define verify_memory_slice(m,i,n) (0)  // PRODUCT version is no-op
+#define verify_memory_slice(m,i,n) (void)(0)  // PRODUCT version is no-op
 #endif
 
 
diff --git a/src/share/vm/prims/forte.cpp b/src/share/vm/prims/forte.cpp
index 737dcecd0..43da74944 100644
--- a/src/share/vm/prims/forte.cpp
+++ b/src/share/vm/prims/forte.cpp
@@ -619,7 +619,7 @@ void    collector_func_load(char* name,
                             void* null_argument_3);
 #pragma weak collector_func_load
 #define collector_func_load(x0,x1,x2,x3,x4,x5,x6) \
-        ( collector_func_load ? collector_func_load(x0,x1,x2,x3,x4,x5,x6),0 : 0 )
+        ( collector_func_load ? collector_func_load(x0,x1,x2,x3,x4,x5,x6),(void)0 : (void)0 )
 #endif // __APPLE__
 #endif // !_WINDOWS
 
diff --git a/src/share/vm/prims/jvm.cpp b/src/share/vm/prims/jvm.cpp
index 08e2f881c..da34f2e13 100644
--- a/src/share/vm/prims/jvm.cpp
+++ b/src/share/vm/prims/jvm.cpp
@@ -3310,24 +3310,10 @@ JVM_ENTRY(jobject, JVM_CurrentClassLoader(JNIEnv *env))
 JVM_END
 
 
-// Utility object for collecting method holders walking down the stack
-class KlassLink: public ResourceObj {
- public:
-  KlassHandle klass;
-  KlassLink*  next;
-
-  KlassLink(KlassHandle k) { klass = k; next = NULL; }
-};
-
-
 JVM_ENTRY(jobjectArray, JVM_GetClassContext(JNIEnv *env))
   JVMWrapper("JVM_GetClassContext");
   ResourceMark rm(THREAD);
   JvmtiVMObjectAllocEventCollector oam;
-  // Collect linked list of (handles to) method holders
-  KlassLink* first = NULL;
-  KlassLink* last  = NULL;
-  int depth = 0;
   vframeStream vfst(thread);
 
   if (SystemDictionary::reflect_CallerSensitive_klass() != NULL) {
@@ -3341,32 +3327,23 @@ JVM_ENTRY(jobjectArray, JVM_GetClassContext(JNIEnv *env))
   }
 
   // Collect method holders
+  GrowableArray<KlassHandle>* klass_array = new GrowableArray<KlassHandle>();
   for (; !vfst.at_end(); vfst.security_next()) {
     Method* m = vfst.method();
     // Native frames are not returned
     if (!m->is_ignored_by_security_stack_walk() && !m->is_native()) {
       Klass* holder = m->method_holder();
       assert(holder->is_klass(), "just checking");
-      depth++;
-      KlassLink* l = new KlassLink(KlassHandle(thread, holder));
-      if (first == NULL) {
-        first = last = l;
-      } else {
-        last->next = l;
-        last = l;
-      }
+      klass_array->append(holder);
     }
   }
 
   // Create result array of type [Ljava/lang/Class;
-  objArrayOop result = oopFactory::new_objArray(SystemDictionary::Class_klass(), depth, CHECK_NULL);
+  objArrayOop result = oopFactory::new_objArray(SystemDictionary::Class_klass(), klass_array->length(), CHECK_NULL);
   // Fill in mirrors corresponding to method holders
-  int index = 0;
-  while (first != NULL) {
-    result->obj_at_put(index++, first->klass()->java_mirror());
-    first = first->next;
+  for (int i = 0; i < klass_array->length(); i++) {
+    result->obj_at_put(i, klass_array->at(i)->java_mirror());
   }
-  assert(index == depth, "just checking");
 
   return (jobjectArray) JNIHandles::make_local(env, result);
 JVM_END
diff --git a/src/share/vm/prims/jvmti.xml b/src/share/vm/prims/jvmti.xml
index dbd6735aa..c23a7b5e1 100644
--- a/src/share/vm/prims/jvmti.xml
+++ b/src/share/vm/prims/jvmti.xml
@@ -1897,7 +1897,7 @@ jvmtiEnv *jvmti;
 	  </description>
 	</param>
 	<param id="monitor_info_ptr">
-	  <allocbuf outcount="owned_monitor_depth_count_ptr">
+	  <allocbuf outcount="monitor_info_count_ptr">
             <struct>jvmtiMonitorStackDepthInfo</struct>
           </allocbuf>
 	  <description>
diff --git a/src/share/vm/prims/whitebox.cpp b/src/share/vm/prims/whitebox.cpp
index 92f2349bb..3552ff206 100644
--- a/src/share/vm/prims/whitebox.cpp
+++ b/src/share/vm/prims/whitebox.cpp
@@ -159,7 +159,7 @@ WB_END
 
 
 WB_ENTRY(void, WB_NMTCommitMemory(JNIEnv* env, jobject o, jlong addr, jlong size))
-  os::commit_memory((char *)(uintptr_t)addr, size);
+  os::commit_memory((char *)(uintptr_t)addr, size, !ExecMem);
   MemTracker::record_virtual_memory_type((address)(uintptr_t)addr, mtTest);
 WB_END
 
diff --git a/src/share/vm/runtime/arguments.cpp b/src/share/vm/runtime/arguments.cpp
index 7cfd7a42e..625feda03 100644
--- a/src/share/vm/runtime/arguments.cpp
+++ b/src/share/vm/runtime/arguments.cpp
@@ -1566,6 +1566,15 @@ julong Arguments::limit_by_allocatable_memory(julong limit) {
   return result;
 }
 
+void Arguments::set_heap_base_min_address() {
+  if (FLAG_IS_DEFAULT(HeapBaseMinAddress) && UseG1GC && HeapBaseMinAddress < 1*G) {
+    // By default HeapBaseMinAddress is 2G on all platforms except Solaris x86.
+    // G1 currently needs a lot of C-heap, so on Solaris we have to give G1
+    // some extra space for the C-heap compared to other collectors.
+    FLAG_SET_ERGO(uintx, HeapBaseMinAddress, 1*G);
+  }
+}
+
 void Arguments::set_heap_size() {
   if (!FLAG_IS_DEFAULT(DefaultMaxRAMFraction)) {
     // Deprecated flag
@@ -1885,21 +1894,6 @@ bool Arguments::check_vm_args_consistency() {
   // Note: Needs platform-dependent factoring.
   bool status = true;
 
-#if ( (defined(COMPILER2) && defined(SPARC)))
-  // NOTE: The call to VM_Version_init depends on the fact that VM_Version_init
-  // on sparc doesn't require generation of a stub as is the case on, e.g.,
-  // x86.  Normally, VM_Version_init must be called from init_globals in
-  // init.cpp, which is called by the initial java thread *after* arguments
-  // have been parsed.  VM_Version_init gets called twice on sparc.
-  extern void VM_Version_init();
-  VM_Version_init();
-  if (!VM_Version::has_v9()) {
-    jio_fprintf(defaultStream::error_stream(),
-                "V8 Machine detected, Server requires V9\n");
-    status = false;
-  }
-#endif /* COMPILER2 && SPARC */
-
   // Allow both -XX:-UseStackBanging and -XX:-UseBoundThreads in non-product
   // builds so the cost of stack banging can be measured.
 #if (defined(PRODUCT) && defined(SOLARIS))
@@ -3525,6 +3519,8 @@ jint Arguments::parse(const JavaVMInitArgs* args) {
     }
   }
 
+  set_heap_base_min_address();
+
   // Set heap size based on available physical memory
   set_heap_size();
 
diff --git a/src/share/vm/runtime/arguments.hpp b/src/share/vm/runtime/arguments.hpp
index 0a4350ee3..0e84208a2 100644
--- a/src/share/vm/runtime/arguments.hpp
+++ b/src/share/vm/runtime/arguments.hpp
@@ -315,6 +315,8 @@ class Arguments : AllStatic {
   // limits the given memory size by the maximum amount of memory this process is
   // currently allowed to allocate or reserve.
   static julong limit_by_allocatable_memory(julong size);
+  // Setup HeapBaseMinAddress
+  static void set_heap_base_min_address();
   // Setup heap size
   static void set_heap_size();
   // Based on automatic selection criteria, should the
diff --git a/src/share/vm/runtime/os.cpp b/src/share/vm/runtime/os.cpp
index 7775d9410..93a60cd06 100644
--- a/src/share/vm/runtime/os.cpp
+++ b/src/share/vm/runtime/os.cpp
@@ -647,10 +647,13 @@ void* os::realloc(void *memblock, size_t size, MEMFLAGS memflags, address caller
 #ifndef ASSERT
   NOT_PRODUCT(inc_stat_counter(&num_mallocs, 1));
   NOT_PRODUCT(inc_stat_counter(&alloc_bytes, size));
+  MemTracker::Tracker tkr = MemTracker::get_realloc_tracker();
   void* ptr = ::realloc(memblock, size);
   if (ptr != NULL) {
-    MemTracker::record_realloc((address)memblock, (address)ptr, size, memflags,
+    tkr.record((address)memblock, (address)ptr, size, memflags,
      caller == 0 ? CALLER_PC : caller);
+  } else {
+    tkr.discard();
   }
   return ptr;
 #else
@@ -1456,7 +1459,7 @@ bool os::create_stack_guard_pages(char* addr, size_t bytes) {
 char* os::reserve_memory(size_t bytes, char* addr, size_t alignment_hint) {
   char* result = pd_reserve_memory(bytes, addr, alignment_hint);
   if (result != NULL) {
-    MemTracker::record_virtual_memory_reserve((address)result, bytes, CALLER_PC);
+    MemTracker::record_virtual_memory_reserve((address)result, bytes, mtNone, CALLER_PC);
   }
 
   return result;
@@ -1466,7 +1469,7 @@ char* os::reserve_memory(size_t bytes, char* addr, size_t alignment_hint,
    MEMFLAGS flags) {
   char* result = pd_reserve_memory(bytes, addr, alignment_hint);
   if (result != NULL) {
-    MemTracker::record_virtual_memory_reserve((address)result, bytes, CALLER_PC);
+    MemTracker::record_virtual_memory_reserve((address)result, bytes, mtNone, CALLER_PC);
     MemTracker::record_virtual_memory_type((address)result, flags);
   }
 
@@ -1476,7 +1479,7 @@ char* os::reserve_memory(size_t bytes, char* addr, size_t alignment_hint,
 char* os::attempt_reserve_memory_at(size_t bytes, char* addr) {
   char* result = pd_attempt_reserve_memory_at(bytes, addr);
   if (result != NULL) {
-    MemTracker::record_virtual_memory_reserve((address)result, bytes, CALLER_PC);
+    MemTracker::record_virtual_memory_reserve((address)result, bytes, mtNone, CALLER_PC);
   }
   return result;
 }
@@ -1503,18 +1506,36 @@ bool os::commit_memory(char* addr, size_t size, size_t alignment_hint,
   return res;
 }
 
+void os::commit_memory_or_exit(char* addr, size_t bytes, bool executable,
+                               const char* mesg) {
+  pd_commit_memory_or_exit(addr, bytes, executable, mesg);
+  MemTracker::record_virtual_memory_commit((address)addr, bytes, CALLER_PC);
+}
+
+void os::commit_memory_or_exit(char* addr, size_t size, size_t alignment_hint,
+                               bool executable, const char* mesg) {
+  os::pd_commit_memory_or_exit(addr, size, alignment_hint, executable, mesg);
+  MemTracker::record_virtual_memory_commit((address)addr, size, CALLER_PC);
+}
+
 bool os::uncommit_memory(char* addr, size_t bytes) {
+  MemTracker::Tracker tkr = MemTracker::get_virtual_memory_uncommit_tracker();
   bool res = pd_uncommit_memory(addr, bytes);
   if (res) {
-    MemTracker::record_virtual_memory_uncommit((address)addr, bytes);
+    tkr.record((address)addr, bytes);
+  } else {
+    tkr.discard();
   }
   return res;
 }
 
 bool os::release_memory(char* addr, size_t bytes) {
+  MemTracker::Tracker tkr = MemTracker::get_virtual_memory_release_tracker();
   bool res = pd_release_memory(addr, bytes);
   if (res) {
-    MemTracker::record_virtual_memory_release((address)addr, bytes);
+    tkr.record((address)addr, bytes);
+  } else {
+    tkr.discard();
   }
   return res;
 }
@@ -1525,8 +1546,7 @@ char* os::map_memory(int fd, const char* file_name, size_t file_offset,
                            bool allow_exec) {
   char* result = pd_map_memory(fd, file_name, file_offset, addr, bytes, read_only, allow_exec);
   if (result != NULL) {
-    MemTracker::record_virtual_memory_reserve((address)result, bytes, CALLER_PC);
-    MemTracker::record_virtual_memory_commit((address)result, bytes, CALLER_PC);
+    MemTracker::record_virtual_memory_reserve_and_commit((address)result, bytes, mtNone, CALLER_PC);
   }
   return result;
 }
@@ -1539,10 +1559,12 @@ char* os::remap_memory(int fd, const char* file_name, size_t file_offset,
 }
 
 bool os::unmap_memory(char *addr, size_t bytes) {
+  MemTracker::Tracker tkr = MemTracker::get_virtual_memory_release_tracker();
   bool result = pd_unmap_memory(addr, bytes);
   if (result) {
-    MemTracker::record_virtual_memory_uncommit((address)addr, bytes);
-    MemTracker::record_virtual_memory_release((address)addr, bytes);
+    tkr.record((address)addr, bytes);
+  } else {
+    tkr.discard();
   }
   return result;
 }
diff --git a/src/share/vm/runtime/os.hpp b/src/share/vm/runtime/os.hpp
index 3f2554399..e1866919d 100644
--- a/src/share/vm/runtime/os.hpp
+++ b/src/share/vm/runtime/os.hpp
@@ -78,6 +78,10 @@ enum ThreadPriority {        // JLS 20.20.1-3
   CriticalPriority = 11      // Critical thread priority
 };
 
+// Executable parameter flag for os::commit_memory() and
+// os::commit_memory_or_exit().
+const bool ExecMem = true;
+
 // Typedef for structured exception handling support
 typedef void (*java_call_t)(JavaValue* value, methodHandle* method, JavaCallArguments* args, Thread* thread);
 
@@ -104,9 +108,16 @@ class os: AllStatic {
   static char*  pd_attempt_reserve_memory_at(size_t bytes, char* addr);
   static void   pd_split_reserved_memory(char *base, size_t size,
                                       size_t split, bool realloc);
-  static bool   pd_commit_memory(char* addr, size_t bytes, bool executable = false);
+  static bool   pd_commit_memory(char* addr, size_t bytes, bool executable);
   static bool   pd_commit_memory(char* addr, size_t size, size_t alignment_hint,
-                              bool executable = false);
+                                 bool executable);
+  // Same as pd_commit_memory() that either succeeds or calls
+  // vm_exit_out_of_memory() with the specified mesg.
+  static void   pd_commit_memory_or_exit(char* addr, size_t bytes,
+                                         bool executable, const char* mesg);
+  static void   pd_commit_memory_or_exit(char* addr, size_t size,
+                                         size_t alignment_hint,
+                                         bool executable, const char* mesg);
   static bool   pd_uncommit_memory(char* addr, size_t bytes);
   static bool   pd_release_memory(char* addr, size_t bytes);
 
@@ -261,9 +272,16 @@ class os: AllStatic {
   static char*  attempt_reserve_memory_at(size_t bytes, char* addr);
   static void   split_reserved_memory(char *base, size_t size,
                                       size_t split, bool realloc);
-  static bool   commit_memory(char* addr, size_t bytes, bool executable = false);
+  static bool   commit_memory(char* addr, size_t bytes, bool executable);
   static bool   commit_memory(char* addr, size_t size, size_t alignment_hint,
-                              bool executable = false);
+                              bool executable);
+  // Same as commit_memory() that either succeeds or calls
+  // vm_exit_out_of_memory() with the specified mesg.
+  static void   commit_memory_or_exit(char* addr, size_t bytes,
+                                      bool executable, const char* mesg);
+  static void   commit_memory_or_exit(char* addr, size_t size,
+                                      size_t alignment_hint,
+                                      bool executable, const char* mesg);
   static bool   uncommit_memory(char* addr, size_t bytes);
   static bool   release_memory(char* addr, size_t bytes);
 
diff --git a/src/share/vm/runtime/sharedRuntime.cpp b/src/share/vm/runtime/sharedRuntime.cpp
index fe55e8b0e..c9f37f4e2 100644
--- a/src/share/vm/runtime/sharedRuntime.cpp
+++ b/src/share/vm/runtime/sharedRuntime.cpp
@@ -2731,7 +2731,7 @@ VMRegPair *SharedRuntime::find_callee_arguments(Symbol* sig, bool has_receiver,
   // ResourceObject, so do not put any ResourceMarks in here.
   char *s = sig->as_C_string();
   int len = (int)strlen(s);
-  *s++; len--;                  // Skip opening paren
+  s++; len--;                   // Skip opening paren
   char *t = s+len;
   while( *(--t) != ')' ) ;      // Find close paren
 
diff --git a/src/share/vm/runtime/virtualspace.cpp b/src/share/vm/runtime/virtualspace.cpp
index ba68e887e..9096a034c 100644
--- a/src/share/vm/runtime/virtualspace.cpp
+++ b/src/share/vm/runtime/virtualspace.cpp
@@ -533,11 +533,13 @@ bool VirtualSpace::expand_by(size_t bytes, bool pre_touch) {
            lower_high() + lower_needs <= lower_high_boundary(),
            "must not expand beyond region");
     if (!os::commit_memory(lower_high(), lower_needs, _executable)) {
-      debug_only(warning("os::commit_memory failed"));
+      debug_only(warning("INFO: os::commit_memory(" PTR_FORMAT
+                         ", lower_needs=" SIZE_FORMAT ", %d) failed",
+                         lower_high(), lower_needs, _executable);)
       return false;
     } else {
       _lower_high += lower_needs;
-     }
+    }
   }
   if (middle_needs > 0) {
     assert(lower_high_boundary() <= middle_high() &&
@@ -545,7 +547,10 @@ bool VirtualSpace::expand_by(size_t bytes, bool pre_touch) {
            "must not expand beyond region");
     if (!os::commit_memory(middle_high(), middle_needs, middle_alignment(),
                            _executable)) {
-      debug_only(warning("os::commit_memory failed"));
+      debug_only(warning("INFO: os::commit_memory(" PTR_FORMAT
+                         ", middle_needs=" SIZE_FORMAT ", " SIZE_FORMAT
+                         ", %d) failed", middle_high(), middle_needs,
+                         middle_alignment(), _executable);)
       return false;
     }
     _middle_high += middle_needs;
@@ -555,7 +560,9 @@ bool VirtualSpace::expand_by(size_t bytes, bool pre_touch) {
            upper_high() + upper_needs <= upper_high_boundary(),
            "must not expand beyond region");
     if (!os::commit_memory(upper_high(), upper_needs, _executable)) {
-      debug_only(warning("os::commit_memory failed"));
+      debug_only(warning("INFO: os::commit_memory(" PTR_FORMAT
+                         ", upper_needs=" SIZE_FORMAT ", %d) failed",
+                         upper_high(), upper_needs, _executable);)
       return false;
     } else {
       _upper_high += upper_needs;
diff --git a/src/share/vm/services/diagnosticArgument.cpp b/src/share/vm/services/diagnosticArgument.cpp
index e3e140532..0a1c20d8a 100644
--- a/src/share/vm/services/diagnosticArgument.cpp
+++ b/src/share/vm/services/diagnosticArgument.cpp
@@ -247,7 +247,7 @@ template <> void DCmdArgument<NanoTimeArgument>::init_value(TRAPS) {
   } else {
     _value._time = 0;
     _value._nanotime = 0;
-    strcmp(_value._unit, "ns");
+    strcpy(_value._unit, "ns");
   }
 }
 
diff --git a/src/share/vm/services/memBaseline.cpp b/src/share/vm/services/memBaseline.cpp
index c2d04106c..21eb8b5d4 100644
--- a/src/share/vm/services/memBaseline.cpp
+++ b/src/share/vm/services/memBaseline.cpp
@@ -130,7 +130,7 @@ bool MemBaseline::baseline_malloc_summary(const MemPointerArray* malloc_records)
       if (malloc_ptr->is_arena_record()) {
         // see if arena memory record present
         MemPointerRecord* next_malloc_ptr = (MemPointerRecordEx*)malloc_itr.peek_next();
-        if (next_malloc_ptr->is_arena_memory_record()) {
+        if (next_malloc_ptr != NULL && next_malloc_ptr->is_arena_memory_record()) {
           assert(next_malloc_ptr->is_memory_record_of_arena(malloc_ptr),
              "Arena records do not match");
           size = next_malloc_ptr->size();
diff --git a/src/share/vm/services/memPtr.hpp b/src/share/vm/services/memPtr.hpp
index d98ca80c4..c54f0934a 100644
--- a/src/share/vm/services/memPtr.hpp
+++ b/src/share/vm/services/memPtr.hpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -457,9 +457,8 @@ class SeqMemPointerRecord : public MemPointerRecord {
  public:
   SeqMemPointerRecord(): _seq(0){ }
 
-  SeqMemPointerRecord(address addr, MEMFLAGS flags, size_t size)
-    : MemPointerRecord(addr, flags, size) {
-    _seq = SequenceGenerator::next();
+  SeqMemPointerRecord(address addr, MEMFLAGS flags, size_t size, jint seq)
+    : MemPointerRecord(addr, flags, size), _seq(seq)  {
   }
 
   SeqMemPointerRecord(const SeqMemPointerRecord& copy_from)
@@ -488,8 +487,8 @@ class SeqMemPointerRecordEx : public MemPointerRecordEx {
   SeqMemPointerRecordEx(): _seq(0) { }
 
   SeqMemPointerRecordEx(address addr, MEMFLAGS flags, size_t size,
-    address pc): MemPointerRecordEx(addr, flags, size, pc) {
-    _seq = SequenceGenerator::next();
+    jint seq, address pc):
+    MemPointerRecordEx(addr, flags, size, pc), _seq(seq)  {
   }
 
   SeqMemPointerRecordEx(const SeqMemPointerRecordEx& copy_from)
diff --git a/src/share/vm/services/memRecorder.cpp b/src/share/vm/services/memRecorder.cpp
index 776ad223c..afe7bd245 100644
--- a/src/share/vm/services/memRecorder.cpp
+++ b/src/share/vm/services/memRecorder.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -69,10 +69,11 @@ MemRecorder::MemRecorder() {
 
   if (_pointer_records != NULL) {
     // recode itself
+    address pc = CURRENT_PC;
     record((address)this, (MemPointerRecord::malloc_tag()|mtNMT|otNMTRecorder),
-        sizeof(MemRecorder), CALLER_PC);
+        sizeof(MemRecorder), SequenceGenerator::next(), pc);
     record((address)_pointer_records, (MemPointerRecord::malloc_tag()|mtNMT|otNMTRecorder),
-        _pointer_records->instance_size(),CURRENT_PC);
+        _pointer_records->instance_size(), SequenceGenerator::next(), pc);
   }
 }
 
@@ -116,7 +117,8 @@ int MemRecorder::sort_record_fn(const void* e1, const void* e2) {
   }
 }
 
-bool MemRecorder::record(address p, MEMFLAGS flags, size_t size, address pc) {
+bool MemRecorder::record(address p, MEMFLAGS flags, size_t size, jint seq, address pc) {
+  assert(seq > 0, "No sequence number");
 #ifdef ASSERT
   if (MemPointerRecord::is_virtual_memory_record(flags)) {
     assert((flags & MemPointerRecord::tag_masks) != 0, "bad virtual memory record");
@@ -133,11 +135,11 @@ bool MemRecorder::record(address p, MEMFLAGS flags, size_t size, address pc) {
 #endif
 
   if (MemTracker::track_callsite()) {
-    SeqMemPointerRecordEx ap(p, flags, size, pc);
+    SeqMemPointerRecordEx ap(p, flags, size, seq, pc);
     debug_only(check_dup_seq(ap.seq());)
     return _pointer_records->append(&ap);
   } else {
-    SeqMemPointerRecord ap(p, flags, size);
+    SeqMemPointerRecord ap(p, flags, size, seq);
     debug_only(check_dup_seq(ap.seq());)
     return _pointer_records->append(&ap);
   }
diff --git a/src/share/vm/services/memRecorder.hpp b/src/share/vm/services/memRecorder.hpp
index 9d6bf2b7e..c52fad72e 100644
--- a/src/share/vm/services/memRecorder.hpp
+++ b/src/share/vm/services/memRecorder.hpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -220,7 +220,7 @@ class MemRecorder : public CHeapObj<mtNMT|otNMTRecorder> {
   ~MemRecorder();
 
   // record a memory operation
-  bool record(address addr, MEMFLAGS flags, size_t size, address caller_pc = 0);
+  bool record(address addr, MEMFLAGS flags, size_t size, jint seq, address caller_pc = 0);
 
   // linked list support
   inline void set_next(MemRecorder* rec) {
diff --git a/src/share/vm/services/memReporter.cpp b/src/share/vm/services/memReporter.cpp
index 0311675f3..573d9e03c 100644
--- a/src/share/vm/services/memReporter.cpp
+++ b/src/share/vm/services/memReporter.cpp
@@ -190,17 +190,18 @@ void BaselineReporter::diff_callsites(const MemBaseline& cur, const MemBaseline&
   while (cur_malloc_callsite != NULL || prev_malloc_callsite != NULL) {
     if (prev_malloc_callsite == NULL ||
         cur_malloc_callsite->addr() < prev_malloc_callsite->addr()) {
+      // this is a new callsite
       _outputer.diff_malloc_callsite(cur_malloc_callsite->addr(),
         amount_in_current_scale(cur_malloc_callsite->amount()),
         cur_malloc_callsite->count(),
         diff_in_current_scale(cur_malloc_callsite->amount(), 0),
         diff(cur_malloc_callsite->count(), 0));
       cur_malloc_callsite = (MallocCallsitePointer*)cur_malloc_itr.next();
-    } else if (prev_malloc_callsite == NULL ||
+    } else if (cur_malloc_callsite == NULL ||
                cur_malloc_callsite->addr() > prev_malloc_callsite->addr()) {
-      _outputer.diff_malloc_callsite(cur_malloc_callsite->addr(),
-        amount_in_current_scale(prev_malloc_callsite->amount()),
-        prev_malloc_callsite->count(),
+      // this callsite is already gone
+      _outputer.diff_malloc_callsite(prev_malloc_callsite->addr(),
+        amount_in_current_scale(0), 0,
         diff_in_current_scale(0, prev_malloc_callsite->amount()),
         diff(0, prev_malloc_callsite->count()));
       prev_malloc_callsite = (MallocCallsitePointer*)prev_malloc_itr.next();
@@ -222,6 +223,7 @@ void BaselineReporter::diff_callsites(const MemBaseline& cur, const MemBaseline&
   VMCallsitePointer*          prev_vm_callsite = (VMCallsitePointer*)prev_vm_itr.current();
   while (cur_vm_callsite != NULL || prev_vm_callsite != NULL) {
     if (prev_vm_callsite == NULL || cur_vm_callsite->addr() < prev_vm_callsite->addr()) {
+      // this is a new callsite
       _outputer.diff_virtual_memory_callsite(cur_vm_callsite->addr(),
         amount_in_current_scale(cur_vm_callsite->reserved_amount()),
         amount_in_current_scale(cur_vm_callsite->committed_amount()),
@@ -229,9 +231,10 @@ void BaselineReporter::diff_callsites(const MemBaseline& cur, const MemBaseline&
         diff_in_current_scale(cur_vm_callsite->committed_amount(), 0));
       cur_vm_callsite = (VMCallsitePointer*)cur_vm_itr.next();
     } else if (cur_vm_callsite == NULL || cur_vm_callsite->addr() > prev_vm_callsite->addr()) {
+      // this callsite is already gone
       _outputer.diff_virtual_memory_callsite(prev_vm_callsite->addr(),
-        amount_in_current_scale(prev_vm_callsite->reserved_amount()),
-        amount_in_current_scale(prev_vm_callsite->committed_amount()),
+        amount_in_current_scale(0),
+        amount_in_current_scale(0),
         diff_in_current_scale(0, prev_vm_callsite->reserved_amount()),
         diff_in_current_scale(0, prev_vm_callsite->committed_amount()));
       prev_vm_callsite = (VMCallsitePointer*)prev_vm_itr.next();
diff --git a/src/share/vm/services/memTracker.cpp b/src/share/vm/services/memTracker.cpp
index 0e6f7c42e..e0a1b29a8 100644
--- a/src/share/vm/services/memTracker.cpp
+++ b/src/share/vm/services/memTracker.cpp
@@ -69,6 +69,7 @@ int                             MemTracker::_thread_count = 255;
 volatile jint                   MemTracker::_pooled_recorder_count = 0;
 volatile unsigned long          MemTracker::_processing_generation = 0;
 volatile bool                   MemTracker::_worker_thread_idle = false;
+volatile jint                   MemTracker::_pending_op_count = 0;
 volatile bool                   MemTracker::_slowdown_calling_thread = false;
 debug_only(intx                 MemTracker::_main_thread_tid = 0;)
 NOT_PRODUCT(volatile jint       MemTracker::_pending_recorder_count = 0;)
@@ -337,92 +338,14 @@ void MemTracker::release_thread_recorder(MemRecorder* rec) {
   Atomic::inc(&_pooled_recorder_count);
 }
 
-/*
- * This is the most important method in whole nmt implementation.
- *
- * Create a memory record.
- * 1. When nmt is in single-threaded bootstrapping mode, no lock is needed as VM
- *    still in single thread mode.
- * 2. For all threads other than JavaThread, ThreadCritical is needed
- *    to write to recorders to global recorder.
- * 3. For JavaThreads that are not longer visible by safepoint, also
- *    need to take ThreadCritical and records are written to global
- *    recorders, since these threads are NOT walked by Threads.do_thread().
- * 4. JavaThreads that are running in native state, have to transition
- *    to VM state before writing to per-thread recorders.
- * 5. JavaThreads that are running in VM state do not need any lock and
- *    records are written to per-thread recorders.
- * 6. For a thread has yet to attach VM 'Thread', they need to take
- *    ThreadCritical to write to global recorder.
- *
- *    Important note:
- *    NO LOCK should be taken inside ThreadCritical lock !!!
- */
-void MemTracker::create_memory_record(address addr, MEMFLAGS flags,
-    size_t size, address pc, Thread* thread) {
-  assert(addr != NULL, "Sanity check");
-  if (!shutdown_in_progress()) {
-    // single thread, we just write records direct to global recorder,'
-    // with any lock
-    if (_state == NMT_bootstrapping_single_thread) {
-      assert(_main_thread_tid == os::current_thread_id(), "wrong thread");
-      thread = NULL;
-    } else {
-      if (thread == NULL) {
-          // don't use Thread::current(), since it is possible that
-          // the calling thread has yet to attach to VM 'Thread',
-          // which will result assertion failure
-          thread = ThreadLocalStorage::thread();
-      }
-    }
-
-    if (thread != NULL) {
-      // slow down all calling threads except NMT worker thread, so it
-      // can catch up.
-      if (_slowdown_calling_thread && thread != _worker_thread) {
-        os::yield_all();
-      }
-
-      if (thread->is_Java_thread() && ((JavaThread*)thread)->is_safepoint_visible()) {
-        JavaThread*      java_thread = (JavaThread*)thread;
-        JavaThreadState  state = java_thread->thread_state();
-        if (SafepointSynchronize::safepoint_safe(java_thread, state)) {
-          // JavaThreads that are safepoint safe, can run through safepoint,
-          // so ThreadCritical is needed to ensure no threads at safepoint create
-          // new records while the records are being gathered and the sequence number is changing
-          ThreadCritical tc;
-          create_record_in_recorder(addr, flags, size, pc, java_thread);
-        } else {
-          create_record_in_recorder(addr, flags, size, pc, java_thread);
-        }
-      } else {
-        // other threads, such as worker and watcher threads, etc. need to
-        // take ThreadCritical to write to global recorder
-        ThreadCritical tc;
-        create_record_in_recorder(addr, flags, size, pc, NULL);
-      }
-    } else {
-      if (_state == NMT_bootstrapping_single_thread) {
-        // single thread, no lock needed
-        create_record_in_recorder(addr, flags, size, pc, NULL);
-      } else {
-        // for thread has yet to attach VM 'Thread', we can not use VM mutex.
-        // use native thread critical instead
-        ThreadCritical tc;
-        create_record_in_recorder(addr, flags, size, pc, NULL);
-      }
-    }
-  }
-}
-
 // write a record to proper recorder. No lock can be taken from this method
 // down.
-void MemTracker::create_record_in_recorder(address addr, MEMFLAGS flags,
-    size_t size, address pc, JavaThread* thread) {
+void MemTracker::write_tracking_record(address addr, MEMFLAGS flags,
+    size_t size, jint seq, address pc, JavaThread* thread) {
 
     MemRecorder* rc = get_thread_recorder(thread);
     if (rc != NULL) {
-      rc->record(addr, flags, size, pc);
+      rc->record(addr, flags, size, seq, pc);
     }
 }
 
@@ -487,39 +410,43 @@ void MemTracker::sync() {
         return;
       }
     }
-    _sync_point_skip_count = 0;
     {
       // This method is running at safepoint, with ThreadCritical lock,
       // it should guarantee that NMT is fully sync-ed.
       ThreadCritical tc;
 
-      SequenceGenerator::reset();
+      // We can NOT execute NMT sync-point if there are pending tracking ops.
+      if (_pending_op_count == 0) {
+        SequenceGenerator::reset();
+        _sync_point_skip_count = 0;
 
-      // walk all JavaThreads to collect recorders
-      SyncThreadRecorderClosure stc;
-      Threads::threads_do(&stc);
+        // walk all JavaThreads to collect recorders
+        SyncThreadRecorderClosure stc;
+        Threads::threads_do(&stc);
 
-      _thread_count = stc.get_thread_count();
-      MemRecorder* pending_recorders = get_pending_recorders();
+        _thread_count = stc.get_thread_count();
+        MemRecorder* pending_recorders = get_pending_recorders();
 
-      if (_global_recorder != NULL) {
-        _global_recorder->set_next(pending_recorders);
-        pending_recorders = _global_recorder;
-        _global_recorder = NULL;
-      }
+        if (_global_recorder != NULL) {
+          _global_recorder->set_next(pending_recorders);
+          pending_recorders = _global_recorder;
+          _global_recorder = NULL;
+        }
 
-      // see if NMT has too many outstanding recorder instances, it usually
-      // means that worker thread is lagging behind in processing them.
-      if (!AutoShutdownNMT) {
-        _slowdown_calling_thread = (MemRecorder::_instance_count > MAX_RECORDER_THREAD_RATIO * _thread_count);
-      }
+        // see if NMT has too many outstanding recorder instances, it usually
+        // means that worker thread is lagging behind in processing them.
+        if (!AutoShutdownNMT) {
+          _slowdown_calling_thread = (MemRecorder::_instance_count > MAX_RECORDER_THREAD_RATIO * _thread_count);
+        }
 
-      // check _worker_thread with lock to avoid racing condition
-      if (_worker_thread != NULL) {
-        _worker_thread->at_sync_point(pending_recorders, InstanceKlass::number_of_instance_classes());
+        // check _worker_thread with lock to avoid racing condition
+        if (_worker_thread != NULL) {
+          _worker_thread->at_sync_point(pending_recorders, InstanceKlass::number_of_instance_classes());
+        }
+        assert(SequenceGenerator::peek() == 1, "Should not have memory activities during sync-point");
+      } else {
+        _sync_point_skip_count ++;
       }
-
-      assert(SequenceGenerator::peek() == 1, "Should not have memory activities during sync-point");
     }
   }
 
@@ -708,3 +635,243 @@ void MemTracker::print_tracker_stats(outputStream* st) {
 }
 #endif
 
+
+// Tracker Implementation
+
+/*
+ * Create a tracker.
+ * This is a fairly complicated constructor, as it has to make two important decisions:
+ *   1) Does it need to take ThreadCritical lock to write tracking record
+ *   2) Does it need to pre-reserve a sequence number for the tracking record
+ *
+ * The rules to determine if ThreadCritical is needed:
+ *   1. When nmt is in single-threaded bootstrapping mode, no lock is needed as VM
+ *      still in single thread mode.
+ *   2. For all threads other than JavaThread, ThreadCritical is needed
+ *      to write to recorders to global recorder.
+ *   3. For JavaThreads that are no longer visible by safepoint, also
+ *      need to take ThreadCritical and records are written to global
+ *      recorders, since these threads are NOT walked by Threads.do_thread().
+ *   4. JavaThreads that are running in safepoint-safe states do not stop
+ *      for safepoints, ThreadCritical lock should be taken to write
+ *      memory records.
+ *   5. JavaThreads that are running in VM state do not need any lock and
+ *      records are written to per-thread recorders.
+ *   6. For a thread has yet to attach VM 'Thread', they need to take
+ *      ThreadCritical to write to global recorder.
+ *
+ *  The memory operations that need pre-reserve sequence numbers:
+ *    The memory operations that "release" memory blocks and the
+ *    operations can fail, need to pre-reserve sequence number. They
+ *    are realloc, uncommit and release.
+ *
+ *  The reason for pre-reserve sequence number, is to prevent race condition:
+ *    Thread 1                      Thread 2
+ *    <release>
+ *                                  <allocate>
+ *                                  <write allocate record>
+ *   <write release record>
+ *   if Thread 2 happens to obtain the memory address Thread 1 just released,
+ *   then NMT can mistakenly report the memory is free.
+ *
+ *  Noticeably, free() does not need pre-reserve sequence number, because the call
+ *  does not fail, so we can alway write "release" record before the memory is actaully
+ *  freed.
+ *
+ *  For realloc, uncommit and release, following coding pattern should be used:
+ *
+ *     MemTracker::Tracker tkr = MemTracker::get_realloc_tracker();
+ *     ptr = ::realloc(...);
+ *     if (ptr == NULL) {
+ *       tkr.record(...)
+ *     } else {
+ *       tkr.discard();
+ *     }
+ *
+ *     MemTracker::Tracker tkr = MemTracker::get_virtual_memory_uncommit_tracker();
+ *     if (uncommit(...)) {
+ *       tkr.record(...);
+ *     } else {
+ *       tkr.discard();
+ *     }
+ *
+ *     MemTracker::Tracker tkr = MemTracker::get_virtual_memory_release_tracker();
+ *     if (release(...)) {
+ *       tkr.record(...);
+ *     } else {
+ *       tkr.discard();
+ *     }
+ *
+ * Since pre-reserved sequence number is only good for the generation that it is acquired,
+ * when there is pending Tracker that reserved sequence number, NMT sync-point has
+ * to be skipped to prevent from advancing generation. This is done by inc and dec
+ * MemTracker::_pending_op_count, when MemTracker::_pending_op_count > 0, NMT sync-point is skipped.
+ * Not all pre-reservation of sequence number will increment pending op count. For JavaThreads
+ * that honor safepoints, safepoint can not occur during the memory operations, so the
+ * pre-reserved sequence number won't cross the generation boundry.
+ */
+MemTracker::Tracker::Tracker(MemoryOperation op, Thread* thr) {
+  _op = NoOp;
+  _seq = 0;
+  if (MemTracker::is_on()) {
+    _java_thread = NULL;
+    _op = op;
+
+    // figure out if ThreadCritical lock is needed to write this operation
+    // to MemTracker
+    if (MemTracker::is_single_threaded_bootstrap()) {
+      thr = NULL;
+    } else if (thr == NULL) {
+      // don't use Thread::current(), since it is possible that
+      // the calling thread has yet to attach to VM 'Thread',
+      // which will result assertion failure
+      thr = ThreadLocalStorage::thread();
+    }
+
+    if (thr != NULL) {
+      // Check NMT load
+      MemTracker::check_NMT_load(thr);
+
+      if (thr->is_Java_thread() && ((JavaThread*)thr)->is_safepoint_visible()) {
+        _java_thread = (JavaThread*)thr;
+        JavaThreadState  state = _java_thread->thread_state();
+        // JavaThreads that are safepoint safe, can run through safepoint,
+        // so ThreadCritical is needed to ensure no threads at safepoint create
+        // new records while the records are being gathered and the sequence number is changing
+        _need_thread_critical_lock =
+          SafepointSynchronize::safepoint_safe(_java_thread, state);
+      } else {
+        _need_thread_critical_lock = true;
+      }
+    } else {
+       _need_thread_critical_lock
+         = !MemTracker::is_single_threaded_bootstrap();
+    }
+
+    // see if we need to pre-reserve sequence number for this operation
+    if (_op == Realloc || _op == Uncommit || _op == Release) {
+      if (_need_thread_critical_lock) {
+        ThreadCritical tc;
+        MemTracker::inc_pending_op_count();
+        _seq = SequenceGenerator::next();
+      } else {
+        // for the threads that honor safepoints, no safepoint can occur
+        // during the lifespan of tracker, so we don't need to increase
+        // pending op count.
+        _seq = SequenceGenerator::next();
+      }
+    }
+  }
+}
+
+void MemTracker::Tracker::discard() {
+  if (MemTracker::is_on() && _seq != 0) {
+    if (_need_thread_critical_lock) {
+      ThreadCritical tc;
+      MemTracker::dec_pending_op_count();
+    }
+    _seq = 0;
+  }
+}
+
+
+void MemTracker::Tracker::record(address old_addr, address new_addr, size_t size,
+  MEMFLAGS flags, address pc) {
+  assert(old_addr != NULL && new_addr != NULL, "Sanity check");
+  assert(_op == Realloc || _op == NoOp, "Wrong call");
+  if (MemTracker::is_on() && NMT_CAN_TRACK(flags) && _op != NoOp) {
+    assert(_seq > 0, "Need pre-reserve sequence number");
+    if (_need_thread_critical_lock) {
+      ThreadCritical tc;
+      // free old address, use pre-reserved sequence number
+      MemTracker::write_tracking_record(old_addr, MemPointerRecord::free_tag(),
+        0, _seq, pc, _java_thread);
+      MemTracker::write_tracking_record(new_addr, flags | MemPointerRecord::malloc_tag(),
+        size, SequenceGenerator::next(), pc, _java_thread);
+      // decrement MemTracker pending_op_count
+      MemTracker::dec_pending_op_count();
+    } else {
+      // free old address, use pre-reserved sequence number
+      MemTracker::write_tracking_record(old_addr, MemPointerRecord::free_tag(),
+        0, _seq, pc, _java_thread);
+      MemTracker::write_tracking_record(new_addr, flags | MemPointerRecord::malloc_tag(),
+        size, SequenceGenerator::next(), pc, _java_thread);
+    }
+    _seq = 0;
+  }
+}
+
+void MemTracker::Tracker::record(address addr, size_t size, MEMFLAGS flags, address pc) {
+  // OOM already?
+  if (addr == NULL) return;
+
+  if (MemTracker::is_on() && NMT_CAN_TRACK(flags) && _op != NoOp) {
+    bool pre_reserved_seq = (_seq != 0);
+    address  pc = CALLER_CALLER_PC;
+    MEMFLAGS orig_flags = flags;
+
+    // or the tagging flags
+    switch(_op) {
+      case Malloc:
+        flags |= MemPointerRecord::malloc_tag();
+        break;
+      case Free:
+        flags = MemPointerRecord::free_tag();
+        break;
+      case Realloc:
+        fatal("Use the other Tracker::record()");
+        break;
+      case Reserve:
+      case ReserveAndCommit:
+        flags |= MemPointerRecord::virtual_memory_reserve_tag();
+        break;
+      case Commit:
+        flags = MemPointerRecord::virtual_memory_commit_tag();
+        break;
+      case Type:
+        flags |= MemPointerRecord::virtual_memory_type_tag();
+        break;
+      case Uncommit:
+        assert(pre_reserved_seq, "Need pre-reserve sequence number");
+        flags = MemPointerRecord::virtual_memory_uncommit_tag();
+        break;
+      case Release:
+        assert(pre_reserved_seq, "Need pre-reserve sequence number");
+        flags = MemPointerRecord::virtual_memory_release_tag();
+        break;
+      case ArenaSize:
+        // a bit of hack here, add a small postive offset to arena
+        // address for its size record, so the size record is sorted
+        // right after arena record.
+        flags = MemPointerRecord::arena_size_tag();
+        addr += sizeof(void*);
+        break;
+      case StackRelease:
+        flags = MemPointerRecord::virtual_memory_release_tag();
+        break;
+      default:
+        ShouldNotReachHere();
+    }
+
+    // write memory tracking record
+    if (_need_thread_critical_lock) {
+      ThreadCritical tc;
+      if (_seq == 0) _seq = SequenceGenerator::next();
+      MemTracker::write_tracking_record(addr, flags, size, _seq, pc, _java_thread);
+      if (_op == ReserveAndCommit) {
+        MemTracker::write_tracking_record(addr, orig_flags | MemPointerRecord::virtual_memory_commit_tag(),
+          size, SequenceGenerator::next(), pc, _java_thread);
+      }
+      if (pre_reserved_seq) MemTracker::dec_pending_op_count();
+    } else {
+      if (_seq == 0) _seq = SequenceGenerator::next();
+      MemTracker::write_tracking_record(addr, flags, size, _seq, pc, _java_thread);
+      if (_op == ReserveAndCommit) {
+        MemTracker::write_tracking_record(addr, orig_flags | MemPointerRecord::virtual_memory_commit_tag(),
+          size, SequenceGenerator::next(), pc, _java_thread);
+      }
+    }
+    _seq = 0;
+  }
+}
+
diff --git a/src/share/vm/services/memTracker.hpp b/src/share/vm/services/memTracker.hpp
index a7d067552..dc7b78859 100644
--- a/src/share/vm/services/memTracker.hpp
+++ b/src/share/vm/services/memTracker.hpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -54,6 +54,18 @@ class MemTracker : AllStatic {
       NMT_sequence_overflow  // overflow the sequence number
    };
 
+  class Tracker {
+   public:
+    void discard() { }
+
+    void record(address addr, size_t size = 0, MEMFLAGS flags = mtNone, address pc = NULL) { }
+    void record(address old_addr, address new_addr, size_t size,
+      MEMFLAGS flags, address pc = NULL) { }
+  };
+
+  private:
+   static Tracker  _tkr;
+
 
   public:
    static inline void init_tracking_options(const char* option_line) { }
@@ -68,19 +80,18 @@ class MemTracker : AllStatic {
    static inline void record_malloc(address addr, size_t size, MEMFLAGS flags,
         address pc = 0, Thread* thread = NULL) { }
    static inline void record_free(address addr, MEMFLAGS flags, Thread* thread = NULL) { }
-   static inline void record_realloc(address old_addr, address new_addr, size_t size,
-        MEMFLAGS flags, address pc = 0, Thread* thread = NULL) { }
    static inline void record_arena_size(address addr, size_t size) { }
    static inline void record_virtual_memory_reserve(address addr, size_t size,
-        address pc = 0, Thread* thread = NULL) { }
+        MEMFLAGS flags, address pc = 0, Thread* thread = NULL) { }
+   static inline void record_virtual_memory_reserve_and_commit(address addr, size_t size,
+        MEMFLAGS flags, address pc = 0, Thread* thread = NULL) { }
    static inline void record_virtual_memory_commit(address addr, size_t size,
         address pc = 0, Thread* thread = NULL) { }
-   static inline void record_virtual_memory_uncommit(address addr, size_t size,
-        Thread* thread = NULL) { }
-   static inline void record_virtual_memory_release(address addr, size_t size,
-        Thread* thread = NULL) { }
    static inline void record_virtual_memory_type(address base, MEMFLAGS flags,
         Thread* thread = NULL) { }
+   static inline Tracker get_realloc_tracker() { return _tkr; }
+   static inline Tracker get_virtual_memory_uncommit_tracker() { return _tkr; }
+   static inline Tracker get_virtual_memory_release_tracker()  { return _tkr; }
    static inline bool baseline() { return false; }
    static inline bool has_baseline() { return false; }
 
@@ -165,6 +176,45 @@ class MemTracker : AllStatic {
   };
 
  public:
+  class Tracker : public StackObj {
+    friend class MemTracker;
+   public:
+    enum MemoryOperation {
+      NoOp,                   // no op
+      Malloc,                 // malloc
+      Realloc,                // realloc
+      Free,                   // free
+      Reserve,                // virtual memory reserve
+      Commit,                 // virtual memory commit
+      ReserveAndCommit,       // virtual memory reserve and commit
+      StackAlloc = ReserveAndCommit, // allocate thread stack
+      Type,                   // assign virtual memory type
+      Uncommit,               // virtual memory uncommit
+      Release,                // virtual memory release
+      ArenaSize,              // set arena size
+      StackRelease            // release thread stack
+    };
+
+
+   protected:
+    Tracker(MemoryOperation op, Thread* thr = NULL);
+
+   public:
+    void discard();
+
+    void record(address addr, size_t size = 0, MEMFLAGS flags = mtNone, address pc = NULL);
+    void record(address old_addr, address new_addr, size_t size,
+      MEMFLAGS flags, address pc = NULL);
+
+   private:
+    bool            _need_thread_critical_lock;
+    JavaThread*     _java_thread;
+    MemoryOperation _op;          // memory operation
+    jint            _seq;         // reserved sequence number
+  };
+
+
+ public:
   // native memory tracking level
   enum NMTLevel {
     NMT_off,              // native memory tracking is off
@@ -276,109 +326,74 @@ class MemTracker : AllStatic {
   // record a 'malloc' call
   static inline void record_malloc(address addr, size_t size, MEMFLAGS flags,
                             address pc = 0, Thread* thread = NULL) {
-    if (is_on() && NMT_CAN_TRACK(flags)) {
-      assert(size > 0, "Sanity check");
-      create_memory_record(addr, (flags|MemPointerRecord::malloc_tag()), size, pc, thread);
-    }
+    Tracker tkr(Tracker::Malloc, thread);
+    tkr.record(addr, size, flags, pc);
   }
   // record a 'free' call
   static inline void record_free(address addr, MEMFLAGS flags, Thread* thread = NULL) {
-    if (is_on() && NMT_CAN_TRACK(flags)) {
-      create_memory_record(addr, MemPointerRecord::free_tag(), 0, 0, thread);
-    }
-  }
-  // record a 'realloc' call
-  static inline void record_realloc(address old_addr, address new_addr, size_t size,
-       MEMFLAGS flags, address pc = 0, Thread* thread = NULL) {
-    if (is_on() && NMT_CAN_TRACK(flags)) {
-      assert(size > 0, "Sanity check");
-      record_free(old_addr, flags, thread);
-      record_malloc(new_addr, size, flags, pc, thread);
-    }
+    Tracker tkr(Tracker::Free, thread);
+    tkr.record(addr, 0, flags, DEBUG_CALLER_PC);
   }
 
-  // record arena memory size
   static inline void record_arena_size(address addr, size_t size) {
-    // we add a positive offset to arena address, so we can have arena memory record
-    // sorted after arena record
-    if (is_on() && !UseMallocOnly) {
-      assert(addr != NULL, "Sanity check");
-      create_memory_record((addr + sizeof(void*)), MemPointerRecord::arena_size_tag(), size,
-        DEBUG_CALLER_PC, NULL);
-    }
+    Tracker tkr(Tracker::ArenaSize);
+    tkr.record(addr, size);
   }
 
   // record a virtual memory 'reserve' call
   static inline void record_virtual_memory_reserve(address addr, size_t size,
-                            address pc = 0, Thread* thread = NULL) {
-    if (is_on()) {
-      assert(size > 0, "Sanity check");
-      create_memory_record(addr, MemPointerRecord::virtual_memory_reserve_tag(),
-                           size, pc, thread);
-    }
+                     MEMFLAGS flags, address pc = 0, Thread* thread = NULL) {
+    assert(size > 0, "Sanity check");
+    Tracker tkr(Tracker::Reserve, thread);
+    tkr.record(addr, size, flags, pc);
   }
 
   static inline void record_thread_stack(address addr, size_t size, Thread* thr,
                            address pc = 0) {
-    if (is_on()) {
-      assert(size > 0 && thr != NULL, "Sanity check");
-      create_memory_record(addr, MemPointerRecord::virtual_memory_reserve_tag() | mtThreadStack,
-                          size, pc, thr);
-      create_memory_record(addr, MemPointerRecord::virtual_memory_commit_tag() | mtThreadStack,
-                          size, pc, thr);
-    }
+    Tracker tkr(Tracker::StackAlloc, thr);
+    tkr.record(addr, size, mtThreadStack, pc);
   }
 
   static inline void release_thread_stack(address addr, size_t size, Thread* thr) {
-    if (is_on()) {
-      assert(size > 0 && thr != NULL, "Sanity check");
-      assert(!thr->is_Java_thread(), "too early");
-      create_memory_record(addr, MemPointerRecord::virtual_memory_uncommit_tag() | mtThreadStack,
-                          size, DEBUG_CALLER_PC, thr);
-      create_memory_record(addr, MemPointerRecord::virtual_memory_release_tag() | mtThreadStack,
-                          size, DEBUG_CALLER_PC, thr);
-    }
+    Tracker tkr(Tracker::StackRelease, thr);
+    tkr.record(addr, size, mtThreadStack, DEBUG_CALLER_PC);
   }
 
   // record a virtual memory 'commit' call
   static inline void record_virtual_memory_commit(address addr, size_t size,
                             address pc, Thread* thread = NULL) {
-    if (is_on()) {
-      assert(size > 0, "Sanity check");
-      create_memory_record(addr, MemPointerRecord::virtual_memory_commit_tag(),
-                           size, pc, thread);
-    }
+    Tracker tkr(Tracker::Commit, thread);
+    tkr.record(addr, size, mtNone, pc);
   }
 
-  // record a virtual memory 'uncommit' call
-  static inline void record_virtual_memory_uncommit(address addr, size_t size,
-                            Thread* thread = NULL) {
-    if (is_on()) {
-      assert(size > 0, "Sanity check");
-      create_memory_record(addr, MemPointerRecord::virtual_memory_uncommit_tag(),
-                           size, DEBUG_CALLER_PC, thread);
-    }
+  static inline void record_virtual_memory_reserve_and_commit(address addr, size_t size,
+    MEMFLAGS flags, address pc, Thread* thread = NULL) {
+    Tracker tkr(Tracker::ReserveAndCommit, thread);
+    tkr.record(addr, size, flags, pc);
   }
 
-  // record a virtual memory 'release' call
-  static inline void record_virtual_memory_release(address addr, size_t size,
-                            Thread* thread = NULL) {
-    if (is_on()) {
-      assert(size > 0, "Sanity check");
-      create_memory_record(addr, MemPointerRecord::virtual_memory_release_tag(),
-                           size, DEBUG_CALLER_PC, thread);
-    }
-  }
 
   // record memory type on virtual memory base address
   static inline void record_virtual_memory_type(address base, MEMFLAGS flags,
                             Thread* thread = NULL) {
-    if (is_on()) {
-      assert(base > 0, "wrong base address");
-      assert((flags & (~mt_masks)) == 0, "memory type only");
-      create_memory_record(base, (flags | MemPointerRecord::virtual_memory_type_tag()),
-                           0, DEBUG_CALLER_PC, thread);
-    }
+    Tracker tkr(Tracker::Type);
+    tkr.record(base, 0, flags);
+  }
+
+  // Get memory trackers for memory operations that can result race conditions.
+  // The memory tracker has to be obtained before realloc, virtual memory uncommit
+  // and virtual memory release, and call tracker.record() method if operation
+  // succeeded, or tracker.discard() to abort the tracking.
+  static inline Tracker get_realloc_tracker() {
+    return Tracker(Tracker::Realloc);
+  }
+
+  static inline Tracker get_virtual_memory_uncommit_tracker() {
+    return Tracker(Tracker::Uncommit);
+  }
+
+  static inline Tracker get_virtual_memory_release_tracker() {
+    return Tracker(Tracker::Release);
   }
 
 
@@ -444,6 +459,31 @@ class MemTracker : AllStatic {
   static MemRecorder* get_pending_recorders();
   static void delete_all_pending_recorders();
 
+  // write a memory tracking record in recorder
+  static void write_tracking_record(address addr, MEMFLAGS type,
+    size_t size, jint seq, address pc, JavaThread* thread);
+
+  static bool is_single_threaded_bootstrap() {
+    return _state == NMT_bootstrapping_single_thread;
+  }
+
+  static void check_NMT_load(Thread* thr) {
+    assert(thr != NULL, "Sanity check");
+    if (_slowdown_calling_thread && thr != _worker_thread) {
+      os::yield_all();
+    }
+  }
+
+  static void inc_pending_op_count() {
+    Atomic::inc(&_pending_op_count);
+  }
+
+  static void dec_pending_op_count() {
+    Atomic::dec(&_pending_op_count);
+    assert(_pending_op_count >= 0, "Sanity check");
+  }
+
+
  private:
   // retrieve a pooled memory record or create new one if there is not
   // one available
@@ -522,6 +562,12 @@ class MemTracker : AllStatic {
   // if NMT should slow down calling thread to allow
   // worker thread to catch up
   static volatile bool             _slowdown_calling_thread;
+
+  // pending memory op count.
+  // Certain memory ops need to pre-reserve sequence number
+  // before memory operation can happen to avoid race condition.
+  // See MemTracker::Tracker for detail
+  static volatile jint             _pending_op_count;
 };
 
 #endif // !INCLUDE_NMT
diff --git a/src/share/vm/services/threadService.cpp b/src/share/vm/services/threadService.cpp
index 03289c7e9..222ae383c 100644
--- a/src/share/vm/services/threadService.cpp
+++ b/src/share/vm/services/threadService.cpp
@@ -327,27 +327,30 @@ DeadlockCycle* ThreadService::find_deadlocks_at_safepoint(bool concurrent_locks)
     while (waitingToLockMonitor != NULL || waitingToLockBlocker != NULL) {
       cycle->add_thread(currentThread);
       if (waitingToLockMonitor != NULL) {
-        currentThread = Threads::owning_thread_from_monitor_owner(
-                          (address)waitingToLockMonitor->owner(),
-                          false /* no locking needed */);
-        if (currentThread == NULL) {
-          // This function is called at a safepoint so the JavaThread
-          // that owns waitingToLockMonitor should be findable, but
-          // if it is not findable, then the previous currentThread is
-          // blocked permanently. We record this as a deadlock.
-          num_deadlocks++;
-
-          cycle->set_deadlock(true);
-
-          // add this cycle to the deadlocks list
-          if (deadlocks == NULL) {
-            deadlocks = cycle;
-          } else {
-            last->set_next(cycle);
+        address currentOwner = (address)waitingToLockMonitor->owner();
+        if (currentOwner != NULL) {
+          currentThread = Threads::owning_thread_from_monitor_owner(
+                            currentOwner,
+                            false /* no locking needed */);
+          if (currentThread == NULL) {
+            // This function is called at a safepoint so the JavaThread
+            // that owns waitingToLockMonitor should be findable, but
+            // if it is not findable, then the previous currentThread is
+            // blocked permanently. We record this as a deadlock.
+            num_deadlocks++;
+
+            cycle->set_deadlock(true);
+
+            // add this cycle to the deadlocks list
+            if (deadlocks == NULL) {
+              deadlocks = cycle;
+            } else {
+              last->set_next(cycle);
+            }
+            last = cycle;
+            cycle = new DeadlockCycle();
+            break;
           }
-          last = cycle;
-          cycle = new DeadlockCycle();
-          break;
         }
       } else {
         if (concurrent_locks) {
diff --git a/src/share/vm/utilities/bitMap.cpp b/src/share/vm/utilities/bitMap.cpp
index 152b40d39..01825d302 100644
--- a/src/share/vm/utilities/bitMap.cpp
+++ b/src/share/vm/utilities/bitMap.cpp
@@ -41,7 +41,7 @@
 
 
 BitMap::BitMap(bm_word_t* map, idx_t size_in_bits) :
-  _map(map), _size(size_in_bits)
+  _map(map), _size(size_in_bits), _map_allocator(false)
 {
   assert(sizeof(bm_word_t) == BytesPerWord, "Implementation assumption.");
   assert(size_in_bits >= 0, "just checking");
@@ -49,7 +49,7 @@ BitMap::BitMap(bm_word_t* map, idx_t size_in_bits) :
 
 
 BitMap::BitMap(idx_t size_in_bits, bool in_resource_area) :
-  _map(NULL), _size(0)
+  _map(NULL), _size(0), _map_allocator(false)
 {
   assert(sizeof(bm_word_t) == BytesPerWord, "Implementation assumption.");
   resize(size_in_bits, in_resource_area);
@@ -65,8 +65,10 @@ void BitMap::resize(idx_t size_in_bits, bool in_resource_area) {
   if (in_resource_area) {
     _map = NEW_RESOURCE_ARRAY(bm_word_t, new_size_in_words);
   } else {
-    if (old_map != NULL) FREE_C_HEAP_ARRAY(bm_word_t, _map, mtInternal);
-    _map = NEW_C_HEAP_ARRAY(bm_word_t, new_size_in_words, mtInternal);
+    if (old_map != NULL) {
+      _map_allocator.free();
+    }
+    _map = _map_allocator.allocate(new_size_in_words);
   }
   Copy::disjoint_words((HeapWord*)old_map, (HeapWord*) _map,
                        MIN2(old_size_in_words, new_size_in_words));
diff --git a/src/share/vm/utilities/bitMap.hpp b/src/share/vm/utilities/bitMap.hpp
index 2486533ab..c4cac118b 100644
--- a/src/share/vm/utilities/bitMap.hpp
+++ b/src/share/vm/utilities/bitMap.hpp
@@ -48,6 +48,7 @@ class BitMap VALUE_OBJ_CLASS_SPEC {
   } RangeSizeHint;
 
  private:
+  ArrayAllocator<bm_word_t, mtInternal> _map_allocator;
   bm_word_t* _map;     // First word in bitmap
   idx_t      _size;    // Size of bitmap (in bits)
 
@@ -113,7 +114,7 @@ class BitMap VALUE_OBJ_CLASS_SPEC {
  public:
 
   // Constructs a bitmap with no map, and size 0.
-  BitMap() : _map(NULL), _size(0) {}
+  BitMap() : _map(NULL), _size(0), _map_allocator(false) {}
 
   // Constructs a bitmap with the given map and size.
   BitMap(bm_word_t* map, idx_t size_in_bits);
diff --git a/src/share/vm/utilities/exceptions.hpp b/src/share/vm/utilities/exceptions.hpp
index 089cd3e08..32c6f35d8 100644
--- a/src/share/vm/utilities/exceptions.hpp
+++ b/src/share/vm/utilities/exceptions.hpp
@@ -194,15 +194,15 @@ class Exceptions {
 #define HAS_PENDING_EXCEPTION                    (((ThreadShadow*)THREAD)->has_pending_exception())
 #define CLEAR_PENDING_EXCEPTION                  (((ThreadShadow*)THREAD)->clear_pending_exception())
 
-#define CHECK                                    THREAD); if (HAS_PENDING_EXCEPTION) return       ; (0
-#define CHECK_(result)                           THREAD); if (HAS_PENDING_EXCEPTION) return result; (0
+#define CHECK                                    THREAD); if (HAS_PENDING_EXCEPTION) return       ; (void)(0
+#define CHECK_(result)                           THREAD); if (HAS_PENDING_EXCEPTION) return result; (void)(0
 #define CHECK_0                                  CHECK_(0)
 #define CHECK_NH                                 CHECK_(Handle())
 #define CHECK_NULL                               CHECK_(NULL)
 #define CHECK_false                              CHECK_(false)
 
-#define CHECK_AND_CLEAR                         THREAD); if (HAS_PENDING_EXCEPTION) { CLEAR_PENDING_EXCEPTION; return;        } (0
-#define CHECK_AND_CLEAR_(result)                THREAD); if (HAS_PENDING_EXCEPTION) { CLEAR_PENDING_EXCEPTION; return result; } (0
+#define CHECK_AND_CLEAR                         THREAD); if (HAS_PENDING_EXCEPTION) { CLEAR_PENDING_EXCEPTION; return;        } (void)(0
+#define CHECK_AND_CLEAR_(result)                THREAD); if (HAS_PENDING_EXCEPTION) { CLEAR_PENDING_EXCEPTION; return result; } (void)(0
 #define CHECK_AND_CLEAR_0                       CHECK_AND_CLEAR_(0)
 #define CHECK_AND_CLEAR_NH                      CHECK_AND_CLEAR_(Handle())
 #define CHECK_AND_CLEAR_NULL                    CHECK_AND_CLEAR_(NULL)
@@ -282,7 +282,7 @@ class Exceptions {
     CLEAR_PENDING_EXCEPTION;               \
     ex->print();                           \
     ShouldNotReachHere();                  \
-  } (0
+  } (void)(0
 
 // ExceptionMark is a stack-allocated helper class for local exception handling.
 // It is used with the EXCEPTION_MARK macro.
diff --git a/src/share/vm/utilities/taskqueue.hpp b/src/share/vm/utilities/taskqueue.hpp
index 980b7e973..aea96c858 100644
--- a/src/share/vm/utilities/taskqueue.hpp
+++ b/src/share/vm/utilities/taskqueue.hpp
@@ -340,8 +340,12 @@ bool GenericTaskQueue<E, F, N>::push_slow(E t, uint dirty_n_elems) {
   if (dirty_n_elems == N - 1) {
     // Actually means 0, so do the push.
     uint localBot = _bottom;
-    // g++ complains if the volatile result of the assignment is unused.
-    const_cast<E&>(_elems[localBot] = t);
+    // g++ complains if the volatile result of the assignment is
+    // unused, so we cast the volatile away.  We cannot cast directly
+    // to void, because gcc treats that as not using the result of the
+    // assignment.  However, casting to E& means that we trigger an
+    // unused-value warning.  So, we cast the E& to void.
+    (void)const_cast<E&>(_elems[localBot] = t);
     OrderAccess::release_store(&_bottom, increment_index(localBot));
     TASKQUEUE_STATS_ONLY(stats.record_push());
     return true;
@@ -397,7 +401,12 @@ bool GenericTaskQueue<E, F, N>::pop_global(E& t) {
     return false;
   }
 
-  const_cast<E&>(t = _elems[oldAge.top()]);
+  // g++ complains if the volatile result of the assignment is
+  // unused, so we cast the volatile away.  We cannot cast directly
+  // to void, because gcc treats that as not using the result of the
+  // assignment.  However, casting to E& means that we trigger an
+  // unused-value warning.  So, we cast the E& to void.
+  (void) const_cast<E&>(t = _elems[oldAge.top()]);
   Age newAge(oldAge);
   newAge.increment();
   Age resAge = _age.cmpxchg(newAge, oldAge);
@@ -640,8 +649,12 @@ GenericTaskQueue<E, F, N>::push(E t) {
   uint dirty_n_elems = dirty_size(localBot, top);
   assert(dirty_n_elems < N, "n_elems out of range.");
   if (dirty_n_elems < max_elems()) {
-    // g++ complains if the volatile result of the assignment is unused.
-    const_cast<E&>(_elems[localBot] = t);
+    // g++ complains if the volatile result of the assignment is
+    // unused, so we cast the volatile away.  We cannot cast directly
+    // to void, because gcc treats that as not using the result of the
+    // assignment.  However, casting to E& means that we trigger an
+    // unused-value warning.  So, we cast the E& to void.
+    (void) const_cast<E&>(_elems[localBot] = t);
     OrderAccess::release_store(&_bottom, increment_index(localBot));
     TASKQUEUE_STATS_ONLY(stats.record_push());
     return true;
@@ -665,7 +678,12 @@ GenericTaskQueue<E, F, N>::pop_local(E& t) {
   // This is necessary to prevent any read below from being reordered
   // before the store just above.
   OrderAccess::fence();
-  const_cast<E&>(t = _elems[localBot]);
+  // g++ complains if the volatile result of the assignment is
+  // unused, so we cast the volatile away.  We cannot cast directly
+  // to void, because gcc treats that as not using the result of the
+  // assignment.  However, casting to E& means that we trigger an
+  // unused-value warning.  So, we cast the E& to void.
+  (void) const_cast<E&>(t = _elems[localBot]);
   // This is a second read of "age"; the "size()" above is the first.
   // If there's still at least one element in the queue, based on the
   // "_bottom" and "age" we've read, then there can be no interference with