diff --git a/src/cpu/x86/vm/c1_Runtime1_x86.cpp b/src/cpu/x86/vm/c1_Runtime1_x86.cpp
index 9810f6e1cdb7bc2644d1671b1c4ddce3663a7a80..aaa79642d7e4f5d13dcb24eb399e3fc64601bf2a 100644
--- a/src/cpu/x86/vm/c1_Runtime1_x86.cpp
+++ b/src/cpu/x86/vm/c1_Runtime1_x86.cpp
@@ -98,7 +98,7 @@ int StubAssembler::call_RT(Register oop_result1, Register metadata_result, addre
   }
   pop(rax);
 #endif
-  reset_last_Java_frame(thread, true, align_stack);
+  reset_last_Java_frame(thread, true);
 
   // discard thread and arguments
   NOT_LP64(addptr(rsp, num_rt_args()*BytesPerWord));
@@ -882,7 +882,7 @@ OopMapSet* Runtime1::generate_patching(StubAssembler* sasm, address target) {
   }
   __ pop(rax);
 #endif
-  __ reset_last_Java_frame(thread, true, false);
+  __ reset_last_Java_frame(thread, true);
 #ifndef _LP64
   __ pop(rcx); // discard thread arg
   __ pop(rcx); // discard dummy
diff --git a/src/cpu/x86/vm/frame_x86.cpp b/src/cpu/x86/vm/frame_x86.cpp
index 1a3364313021a79315744bdfd3212d1ffbed7d87..4cadc43cacbfe451907581889b9433799a611a88 100644
--- a/src/cpu/x86/vm/frame_x86.cpp
+++ b/src/cpu/x86/vm/frame_x86.cpp
@@ -370,13 +370,16 @@ frame frame::sender_for_entry_frame(RegisterMap* map) const {
   JavaFrameAnchor* jfa = entry_frame_call_wrapper()->anchor();
   assert(!entry_frame_is_first(), "next Java fp must be non zero");
   assert(jfa->last_Java_sp() > sp(), "must be above this frame on stack");
+  // Since we are walking the stack now this nested anchor is obviously walkable
+  // even if it wasn't when it was stacked.
+  if (!jfa->walkable()) {
+    // Capture _last_Java_pc (if needed) and mark anchor walkable.
+    jfa->capture_last_Java_pc();
+  }
   map->clear();
   assert(map->include_argument_oops(), "should be set by clear");
-  if (jfa->last_Java_pc() != NULL ) {
-    frame fr(jfa->last_Java_sp(), jfa->last_Java_fp(), jfa->last_Java_pc());
-    return fr;
-  }
-  frame fr(jfa->last_Java_sp(), jfa->last_Java_fp());
+  assert(jfa->last_Java_pc() != NULL, "not walkable");
+  frame fr(jfa->last_Java_sp(), jfa->last_Java_fp(), jfa->last_Java_pc());
   return fr;
 }
 
@@ -714,3 +717,21 @@ frame::frame(void* sp, void* fp, void* pc) {
   init((intptr_t*)sp, (intptr_t*)fp, (address)pc);
 }
 #endif
+
+void JavaFrameAnchor::make_walkable(JavaThread* thread) {
+  // last frame set?
+  if (last_Java_sp() == NULL) return;
+  // already walkable?
+  if (walkable()) return;
+  assert(Thread::current() == (Thread*)thread, "not current thread");
+  assert(last_Java_sp() != NULL, "not called from Java code?");
+  assert(last_Java_pc() == NULL, "already walkable");
+  capture_last_Java_pc();
+  assert(walkable(), "something went wrong");
+}
+
+void JavaFrameAnchor::capture_last_Java_pc() {
+  assert(_last_Java_sp != NULL, "no last frame set");
+  assert(_last_Java_pc == NULL, "already walkable");
+  _last_Java_pc = (address)_last_Java_sp[-1];
+}
diff --git a/src/cpu/x86/vm/frame_x86.inline.hpp b/src/cpu/x86/vm/frame_x86.inline.hpp
index a2649fdb1746168700f61eb44fa8e73927b3eb68..325f780089c755ed4b51c23e73cbf0ff752b0b22 100644
--- a/src/cpu/x86/vm/frame_x86.inline.hpp
+++ b/src/cpu/x86/vm/frame_x86.inline.hpp
@@ -96,6 +96,7 @@ inline frame::frame(intptr_t* sp, intptr_t* fp) {
   // call a specialized frame constructor instead of this one.
   // Then we could use the assert below. However this assert is of somewhat dubious
   // value.
+  // UPDATE: this constructor is only used by trace_method_handle_stub() now.
   // assert(_pc != NULL, "no pc?");
 
   _cb = CodeCache::find_blob(_pc);
diff --git a/src/cpu/x86/vm/javaFrameAnchor_x86.hpp b/src/cpu/x86/vm/javaFrameAnchor_x86.hpp
index 77298e53770013f3482b1c5cde9f02f6655ce02c..40eedf019ae5b19225880ee0b3546b49ea130e2c 100644
--- a/src/cpu/x86/vm/javaFrameAnchor_x86.hpp
+++ b/src/cpu/x86/vm/javaFrameAnchor_x86.hpp
@@ -62,10 +62,9 @@ public:
     _last_Java_sp = src->_last_Java_sp;
   }
 
-  // Always walkable
-  bool walkable(void) { return true; }
-  // Never any thing to do since we are always walkable and can find address of return addresses
-  void make_walkable(JavaThread* thread) { }
+  bool walkable(void)                            { return _last_Java_sp != NULL && _last_Java_pc != NULL; }
+  void make_walkable(JavaThread* thread);
+  void capture_last_Java_pc(void);
 
   intptr_t* last_Java_sp(void) const             { return _last_Java_sp; }
 
diff --git a/src/cpu/x86/vm/macroAssembler_x86.cpp b/src/cpu/x86/vm/macroAssembler_x86.cpp
index 30490d451f84e0465bf507dadf4ea89d786f4cbd..9711089a4dac8fcd997097217a0fb54564de2f40 100644
--- a/src/cpu/x86/vm/macroAssembler_x86.cpp
+++ b/src/cpu/x86/vm/macroAssembler_x86.cpp
@@ -748,8 +748,7 @@ void MacroAssembler::pushptr(AddressLiteral src) {
   }
 }
 
-void MacroAssembler::reset_last_Java_frame(bool clear_fp,
-                                           bool clear_pc) {
+void MacroAssembler::reset_last_Java_frame(bool clear_fp) {
   // we must set sp to zero to clear frame
   movptr(Address(r15_thread, JavaThread::last_Java_sp_offset()), NULL_WORD);
   // must clear fp, so that compiled frames are not confused; it is
@@ -758,9 +757,8 @@ void MacroAssembler::reset_last_Java_frame(bool clear_fp,
     movptr(Address(r15_thread, JavaThread::last_Java_fp_offset()), NULL_WORD);
   }
 
-  if (clear_pc) {
-    movptr(Address(r15_thread, JavaThread::last_Java_pc_offset()), NULL_WORD);
-  }
+  // Always clear the pc because it could have been set by make_walkable()
+  movptr(Address(r15_thread, JavaThread::last_Java_pc_offset()), NULL_WORD);
 }
 
 void MacroAssembler::set_last_Java_frame(Register last_java_sp,
@@ -2561,7 +2559,7 @@ void MacroAssembler::call_VM_base(Register oop_result,
   }
   // reset last Java frame
   // Only interpreter should have to clear fp
-  reset_last_Java_frame(java_thread, true, false);
+  reset_last_Java_frame(java_thread, true);
 
 #ifndef CC_INTERP
    // C++ interp handles this in the interpreter
@@ -3808,7 +3806,7 @@ void MacroAssembler::push_IU_state() {
   pusha();
 }
 
-void MacroAssembler::reset_last_Java_frame(Register java_thread, bool clear_fp, bool clear_pc) {
+void MacroAssembler::reset_last_Java_frame(Register java_thread, bool clear_fp) {
   // determine java_thread register
   if (!java_thread->is_valid()) {
     java_thread = rdi;
@@ -3820,8 +3818,8 @@ void MacroAssembler::reset_last_Java_frame(Register java_thread, bool clear_fp,
     movptr(Address(java_thread, JavaThread::last_Java_fp_offset()), NULL_WORD);
   }
 
-  if (clear_pc)
-    movptr(Address(java_thread, JavaThread::last_Java_pc_offset()), NULL_WORD);
+  // Always clear the pc because it could have been set by make_walkable()
+  movptr(Address(java_thread, JavaThread::last_Java_pc_offset()), NULL_WORD);
 
 }
 
diff --git a/src/cpu/x86/vm/macroAssembler_x86.hpp b/src/cpu/x86/vm/macroAssembler_x86.hpp
index cc5e5856a5efa45de06e0c68c6a6ca67b681da3a..b8270562625fa09ebe3a04b1938d7ea9998b7485 100644
--- a/src/cpu/x86/vm/macroAssembler_x86.hpp
+++ b/src/cpu/x86/vm/macroAssembler_x86.hpp
@@ -289,10 +289,10 @@ class MacroAssembler: public Assembler {
                            Register last_java_fp,
                            address last_java_pc);
 
-  void reset_last_Java_frame(Register thread, bool clear_fp, bool clear_pc);
+  void reset_last_Java_frame(Register thread, bool clear_fp);
 
   // thread in the default location (r15_thread on 64bit)
-  void reset_last_Java_frame(bool clear_fp, bool clear_pc);
+  void reset_last_Java_frame(bool clear_fp);
 
   // Stores
   void store_check(Register obj);                // store check for obj - register is destroyed afterwards
diff --git a/src/cpu/x86/vm/runtime_x86_32.cpp b/src/cpu/x86/vm/runtime_x86_32.cpp
index 36457cb8178014df7b5d3bf4b83d3334ae9af49a..45110b6a0837021b98a5b0c8384f147bc6611574 100644
--- a/src/cpu/x86/vm/runtime_x86_32.cpp
+++ b/src/cpu/x86/vm/runtime_x86_32.cpp
@@ -116,7 +116,7 @@ void OptoRuntime::generate_exception_blob() {
   // No registers to map, rbp is known implicitly
   oop_maps->add_gc_map( __ pc() - start,  new OopMap( framesize, 0 ));
   __ get_thread(rcx);
-  __ reset_last_Java_frame(rcx, false, false);
+  __ reset_last_Java_frame(rcx, false);
 
   // Restore callee-saved registers
   __ movptr(rbp, Address(rsp, rbp_off * wordSize));
diff --git a/src/cpu/x86/vm/sharedRuntime_x86_32.cpp b/src/cpu/x86/vm/sharedRuntime_x86_32.cpp
index c97b43097823d8b8873a1ab536ad515ca0c72763..34418682b9ffb898d1681403587c830bef83a4ef 100644
--- a/src/cpu/x86/vm/sharedRuntime_x86_32.cpp
+++ b/src/cpu/x86/vm/sharedRuntime_x86_32.cpp
@@ -1333,7 +1333,7 @@ static void check_needs_gc_for_critical_native(MacroAssembler* masm,
   __ increment(rsp, wordSize);
 
   __ get_thread(thread);
-  __ reset_last_Java_frame(thread, false, true);
+  __ reset_last_Java_frame(thread, false);
 
   save_or_restore_arguments(masm, stack_slots, total_in_args,
                             arg_save_area, NULL, in_regs, in_sig_bt);
@@ -2251,7 +2251,7 @@ nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
 
   // We can finally stop using that last_Java_frame we setup ages ago
 
-  __ reset_last_Java_frame(thread, false, true);
+  __ reset_last_Java_frame(thread, false);
 
   // Unpack oop result
   if (ret_type == T_OBJECT || ret_type == T_ARRAY) {
@@ -2951,7 +2951,7 @@ void SharedRuntime::generate_deopt_blob() {
   __ pop(rcx);
 
   __ get_thread(rcx);
-  __ reset_last_Java_frame(rcx, false, false);
+  __ reset_last_Java_frame(rcx, false);
 
   // Load UnrollBlock into EDI
   __ mov(rdi, rax);
@@ -3117,7 +3117,7 @@ void SharedRuntime::generate_deopt_blob() {
   __ push(rax);
 
   __ get_thread(rcx);
-  __ reset_last_Java_frame(rcx, false, false);
+  __ reset_last_Java_frame(rcx, false);
 
   // Collect return values
   __ movptr(rax,Address(rsp, (RegisterSaver::raxOffset() + additional_words + 1)*wordSize));
@@ -3219,7 +3219,7 @@ void SharedRuntime::generate_uncommon_trap_blob() {
 
   __ get_thread(rcx);
 
-  __ reset_last_Java_frame(rcx, false, false);
+  __ reset_last_Java_frame(rcx, false);
 
   // Load UnrollBlock into EDI
   __ movptr(rdi, rax);
@@ -3331,7 +3331,7 @@ void SharedRuntime::generate_uncommon_trap_blob() {
   oop_maps->add_gc_map( __ pc()-start, new OopMap( framesize, 0 ) );
 
   __ get_thread(rdi);
-  __ reset_last_Java_frame(rdi, true, false);
+  __ reset_last_Java_frame(rdi, true);
 
   // Pop self-frame.
   __ leave();     // Epilog!
@@ -3426,7 +3426,7 @@ SafepointBlob* SharedRuntime::generate_handler_blob(address call_ptr, int poll_t
 
   // Clear last_Java_sp again
   __ get_thread(java_thread);
-  __ reset_last_Java_frame(java_thread, false, false);
+  __ reset_last_Java_frame(java_thread, false);
 
   __ cmpptr(Address(java_thread, Thread::pending_exception_offset()), (int32_t)NULL_WORD);
   __ jcc(Assembler::equal, noException);
@@ -3501,7 +3501,7 @@ RuntimeStub* SharedRuntime::generate_resolve_blob(address destination, const cha
   __ addptr(rsp, wordSize);
 
   // clear last_Java_sp
-  __ reset_last_Java_frame(thread, true, false);
+  __ reset_last_Java_frame(thread, true);
   // check for pending exceptions
   Label pending;
   __ cmpptr(Address(thread, Thread::pending_exception_offset()), (int32_t)NULL_WORD);
diff --git a/src/cpu/x86/vm/sharedRuntime_x86_64.cpp b/src/cpu/x86/vm/sharedRuntime_x86_64.cpp
index a3eb504c0ce16b872704e86f7f5681bf21eaab69..39a880b5e58d5fe94c2ccd7829188320183a13da 100644
--- a/src/cpu/x86/vm/sharedRuntime_x86_64.cpp
+++ b/src/cpu/x86/vm/sharedRuntime_x86_64.cpp
@@ -1388,7 +1388,7 @@ static void check_needs_gc_for_critical_native(MacroAssembler* masm,
   __ mov(rsp, r12); // restore sp
   __ reinit_heapbase();
 
-  __ reset_last_Java_frame(false, true);
+  __ reset_last_Java_frame(false);
 
   save_or_restore_arguments(masm, stack_slots, total_in_args,
                             arg_save_area, NULL, in_regs, in_sig_bt);
@@ -2497,7 +2497,7 @@ nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
     restore_native_result(masm, ret_type, stack_slots);
   }
 
-  __ reset_last_Java_frame(false, true);
+  __ reset_last_Java_frame(false);
 
   // Unpack oop result
   if (ret_type == T_OBJECT || ret_type == T_ARRAY) {
@@ -3435,7 +3435,7 @@ void SharedRuntime::generate_deopt_blob() {
   // find any register it might need.
   oop_maps->add_gc_map(__ pc() - start, map);
 
-  __ reset_last_Java_frame(false, false);
+  __ reset_last_Java_frame(false);
 
   // Load UnrollBlock* into rdi
   __ mov(rdi, rax);
@@ -3592,7 +3592,7 @@ void SharedRuntime::generate_deopt_blob() {
                        new OopMap( frame_size_in_words, 0 ));
 
   // Clear fp AND pc
-  __ reset_last_Java_frame(true, true);
+  __ reset_last_Java_frame(true);
 
   // Collect return values
   __ movdbl(xmm0, Address(rsp, RegisterSaver::xmm0_offset_in_bytes()));
@@ -3662,7 +3662,7 @@ void SharedRuntime::generate_uncommon_trap_blob() {
 
   oop_maps->add_gc_map(__ pc() - start, map);
 
-  __ reset_last_Java_frame(false, false);
+  __ reset_last_Java_frame(false);
 
   // Load UnrollBlock* into rdi
   __ mov(rdi, rax);
@@ -3775,7 +3775,7 @@ void SharedRuntime::generate_uncommon_trap_blob() {
   oop_maps->add_gc_map(the_pc - start, new OopMap(SimpleRuntimeFrame::framesize, 0));
 
   // Clear fp AND pc
-  __ reset_last_Java_frame(true, true);
+  __ reset_last_Java_frame(true);
 
   // Pop self-frame.
   __ leave();                 // Epilog
@@ -3858,7 +3858,7 @@ SafepointBlob* SharedRuntime::generate_handler_blob(address call_ptr, int poll_t
 
   Label noException;
 
-  __ reset_last_Java_frame(false, false);
+  __ reset_last_Java_frame(false);
 
   __ cmpptr(Address(r15_thread, Thread::pending_exception_offset()), (int32_t)NULL_WORD);
   __ jcc(Assembler::equal, noException);
@@ -3928,7 +3928,7 @@ RuntimeStub* SharedRuntime::generate_resolve_blob(address destination, const cha
   // rax contains the address we are going to jump to assuming no exception got installed
 
   // clear last_Java_sp
-  __ reset_last_Java_frame(false, false);
+  __ reset_last_Java_frame(false);
   // check for pending exceptions
   Label pending;
   __ cmpptr(Address(r15_thread, Thread::pending_exception_offset()), (int32_t)NULL_WORD);
@@ -4309,7 +4309,7 @@ void OptoRuntime::generate_exception_blob() {
 
   oop_maps->add_gc_map(the_pc - start, new OopMap(SimpleRuntimeFrame::framesize, 0));
 
-  __ reset_last_Java_frame(false, true);
+  __ reset_last_Java_frame(false);
 
   // Restore callee-saved registers
 
diff --git a/src/cpu/x86/vm/stubGenerator_x86_32.cpp b/src/cpu/x86/vm/stubGenerator_x86_32.cpp
index 5387ba1842926356a81726930ab9d895d6db0651..50a06d7a5b594915344ebdaff6a992662b49b720 100644
--- a/src/cpu/x86/vm/stubGenerator_x86_32.cpp
+++ b/src/cpu/x86/vm/stubGenerator_x86_32.cpp
@@ -2901,7 +2901,7 @@ class StubGenerator: public StubCodeGenerator {
     // however can use the register value directly if it is callee saved.
     __ get_thread(java_thread);
 
-    __ reset_last_Java_frame(java_thread, true, false);
+    __ reset_last_Java_frame(java_thread, true);
 
     __ leave(); // required for proper stackwalking of RuntimeStub frame
 
diff --git a/src/cpu/x86/vm/stubGenerator_x86_64.cpp b/src/cpu/x86/vm/stubGenerator_x86_64.cpp
index 01fe4b8c5421a15aa674edcafb873a86e2abd502..1d38af79974082248eb3425c0ccc6bd695610235 100644
--- a/src/cpu/x86/vm/stubGenerator_x86_64.cpp
+++ b/src/cpu/x86/vm/stubGenerator_x86_64.cpp
@@ -3923,7 +3923,7 @@ class StubGenerator: public StubCodeGenerator {
 
     oop_maps->add_gc_map(the_pc - start, map);
 
-    __ reset_last_Java_frame(true, true);
+    __ reset_last_Java_frame(true);
 
     __ leave(); // required for proper stackwalking of RuntimeStub frame
 
diff --git a/src/cpu/x86/vm/templateInterpreter_x86_32.cpp b/src/cpu/x86/vm/templateInterpreter_x86_32.cpp
index dfe431a7b7d05df7141593e4bc58017381fb2e3a..4d94902eed1e06868866743209bc3ce9fa8faecd 100644
--- a/src/cpu/x86/vm/templateInterpreter_x86_32.cpp
+++ b/src/cpu/x86/vm/templateInterpreter_x86_32.cpp
@@ -1289,7 +1289,7 @@ address InterpreterGenerator::generate_native_entry(bool synchronized) {
   // change thread state
   __ movl(Address(thread, JavaThread::thread_state_offset()), _thread_in_Java);
 
-  __ reset_last_Java_frame(thread, true, true);
+  __ reset_last_Java_frame(thread, true);
 
   // reset handle block
   __ movptr(t, Address(thread, JavaThread::active_handles_offset()));
@@ -1819,7 +1819,7 @@ void TemplateInterpreterGenerator::generate_throw_exception() {
   __ set_last_Java_frame(thread, noreg, rbp, __ pc());
   __ super_call_VM_leaf(CAST_FROM_FN_PTR(address, InterpreterRuntime::popframe_move_outgoing_args), thread, rax, rbx);
   __ get_thread(thread);
-  __ reset_last_Java_frame(thread, true, true);
+  __ reset_last_Java_frame(thread, true);
   // Restore the last_sp and null it out
   __ movptr(rsp, Address(rbp, frame::interpreter_frame_last_sp_offset * wordSize));
   __ movptr(Address(rbp, frame::interpreter_frame_last_sp_offset * wordSize), NULL_WORD);
diff --git a/src/cpu/x86/vm/templateInterpreter_x86_64.cpp b/src/cpu/x86/vm/templateInterpreter_x86_64.cpp
index 156cbe95d2fe0bbc55fdaee7108443196607dcc9..12173829220b58871682352c75851dd66dd988d7 100644
--- a/src/cpu/x86/vm/templateInterpreter_x86_64.cpp
+++ b/src/cpu/x86/vm/templateInterpreter_x86_64.cpp
@@ -1262,7 +1262,7 @@ address InterpreterGenerator::generate_native_entry(bool synchronized) {
   __ movl(Address(r15_thread, JavaThread::thread_state_offset()), _thread_in_Java);
 
   // reset_last_Java_frame
-  __ reset_last_Java_frame(true, true);
+  __ reset_last_Java_frame(r15_thread, true);
 
   // reset handle block
   __ movptr(t, Address(r15_thread, JavaThread::active_handles_offset()));
@@ -1837,7 +1837,7 @@ void TemplateInterpreterGenerator::generate_throw_exception() {
   // PC must point into interpreter here
   __ set_last_Java_frame(noreg, rbp, __ pc());
   __ super_call_VM_leaf(CAST_FROM_FN_PTR(address, InterpreterRuntime::popframe_move_outgoing_args), r15_thread, c_rarg1, c_rarg2);
-  __ reset_last_Java_frame(true, true);
+  __ reset_last_Java_frame(r15_thread, true);
   // Restore the last_sp and null it out
   __ movptr(rsp, Address(rbp, frame::interpreter_frame_last_sp_offset * wordSize));
   __ movptr(Address(rbp, frame::interpreter_frame_last_sp_offset * wordSize), (int32_t)NULL_WORD);
diff --git a/src/os/linux/vm/os_linux.cpp b/src/os/linux/vm/os_linux.cpp
index 4a4cad5842bf6632091e2d2225872e17d1b7326a..044a70a6dde2ee76370d2e22d1b424a3d2909e9d 100644
--- a/src/os/linux/vm/os_linux.cpp
+++ b/src/os/linux/vm/os_linux.cpp
@@ -2819,11 +2819,8 @@ extern "C" JNIEXPORT void numa_warn(int number, char *where, ...) { }
 extern "C" JNIEXPORT void numa_error(char *where) { }
 extern "C" JNIEXPORT int fork1() { return fork(); }
 
-
-// If we are running with libnuma version > 2, then we should
-// be trying to use symbols with versions 1.1
-// If we are running with earlier version, which did not have symbol versions,
-// we should use the base version.
+// Handle request to load libnuma symbol version 1.1 (API v1). If it fails
+// load symbol from base version instead.
 void* os::Linux::libnuma_dlsym(void* handle, const char *name) {
   void *f = dlvsym(handle, name, "libnuma_1.1");
   if (f == NULL) {
@@ -2832,6 +2829,12 @@ void* os::Linux::libnuma_dlsym(void* handle, const char *name) {
   return f;
 }
 
+// Handle request to load libnuma symbol version 1.2 (API v2) only.
+// Return NULL if the symbol is not defined in this particular version.
+void* os::Linux::libnuma_v2_dlsym(void* handle, const char* name) {
+  return dlvsym(handle, name, "libnuma_1.2");
+}
+
 bool os::Linux::libnuma_init() {
   // sched_getcpu() should be in libc.
   set_sched_getcpu(CAST_TO_FN_PTR(sched_getcpu_func_t,
@@ -2856,6 +2859,8 @@ bool os::Linux::libnuma_init() {
                                             libnuma_dlsym(handle, "numa_tonode_memory")));
       set_numa_interleave_memory(CAST_TO_FN_PTR(numa_interleave_memory_func_t,
                                                 libnuma_dlsym(handle, "numa_interleave_memory")));
+      set_numa_interleave_memory_v2(CAST_TO_FN_PTR(numa_interleave_memory_v2_func_t,
+                                                libnuma_v2_dlsym(handle, "numa_interleave_memory")));
       set_numa_set_bind_policy(CAST_TO_FN_PTR(numa_set_bind_policy_func_t,
                                               libnuma_dlsym(handle, "numa_set_bind_policy")));
       set_numa_bitmask_isbitset(CAST_TO_FN_PTR(numa_bitmask_isbitset_func_t,
@@ -2975,6 +2980,7 @@ os::Linux::numa_num_configured_nodes_func_t os::Linux::_numa_num_configured_node
 os::Linux::numa_available_func_t os::Linux::_numa_available;
 os::Linux::numa_tonode_memory_func_t os::Linux::_numa_tonode_memory;
 os::Linux::numa_interleave_memory_func_t os::Linux::_numa_interleave_memory;
+os::Linux::numa_interleave_memory_v2_func_t os::Linux::_numa_interleave_memory_v2;
 os::Linux::numa_set_bind_policy_func_t os::Linux::_numa_set_bind_policy;
 os::Linux::numa_bitmask_isbitset_func_t os::Linux::_numa_bitmask_isbitset;
 os::Linux::numa_distance_func_t os::Linux::_numa_distance;
diff --git a/src/os/linux/vm/os_linux.hpp b/src/os/linux/vm/os_linux.hpp
index b8c93433e521991d3d706ddd6d3ad5b7fb57ebc9..c56983cdd1a5bcf632b28edc92016dc942b7a577 100644
--- a/src/os/linux/vm/os_linux.hpp
+++ b/src/os/linux/vm/os_linux.hpp
@@ -190,6 +190,8 @@ class Linux {
   static void libpthread_init();
   static bool libnuma_init();
   static void* libnuma_dlsym(void* handle, const char* name);
+  // libnuma v2 (libnuma_1.2) symbols
+  static void* libnuma_v2_dlsym(void* handle, const char* name);
   // Minimum stack size a thread can be created with (allowing
   // the VM to completely create the thread and enter user code)
   static size_t min_stack_allowed;
@@ -250,6 +252,8 @@ private:
   typedef int (*numa_available_func_t)(void);
   typedef int (*numa_tonode_memory_func_t)(void *start, size_t size, int node);
   typedef void (*numa_interleave_memory_func_t)(void *start, size_t size, unsigned long *nodemask);
+  typedef void (*numa_interleave_memory_v2_func_t)(void *start, size_t size, struct bitmask* mask);
+
   typedef void (*numa_set_bind_policy_func_t)(int policy);
   typedef int (*numa_bitmask_isbitset_func_t)(struct bitmask *bmp, unsigned int n);
   typedef int (*numa_distance_func_t)(int node1, int node2);
@@ -261,6 +265,7 @@ private:
   static numa_available_func_t _numa_available;
   static numa_tonode_memory_func_t _numa_tonode_memory;
   static numa_interleave_memory_func_t _numa_interleave_memory;
+  static numa_interleave_memory_v2_func_t _numa_interleave_memory_v2;
   static numa_set_bind_policy_func_t _numa_set_bind_policy;
   static numa_bitmask_isbitset_func_t _numa_bitmask_isbitset;
   static numa_distance_func_t _numa_distance;
@@ -275,6 +280,7 @@ private:
   static void set_numa_available(numa_available_func_t func) { _numa_available = func; }
   static void set_numa_tonode_memory(numa_tonode_memory_func_t func) { _numa_tonode_memory = func; }
   static void set_numa_interleave_memory(numa_interleave_memory_func_t func) { _numa_interleave_memory = func; }
+  static void set_numa_interleave_memory_v2(numa_interleave_memory_v2_func_t func) { _numa_interleave_memory_v2 = func; }
   static void set_numa_set_bind_policy(numa_set_bind_policy_func_t func) { _numa_set_bind_policy = func; }
   static void set_numa_bitmask_isbitset(numa_bitmask_isbitset_func_t func) { _numa_bitmask_isbitset = func; }
   static void set_numa_distance(numa_distance_func_t func) { _numa_distance = func; }
@@ -296,7 +302,10 @@ public:
     return _numa_tonode_memory != NULL ? _numa_tonode_memory(start, size, node) : -1;
   }
   static void numa_interleave_memory(void *start, size_t size) {
-    if (_numa_interleave_memory != NULL && _numa_all_nodes != NULL) {
+    // Use v2 api if available
+    if (_numa_interleave_memory_v2 != NULL && _numa_all_nodes_ptr != NULL) {
+      _numa_interleave_memory_v2(start, size, _numa_all_nodes_ptr);
+    } else if (_numa_interleave_memory != NULL && _numa_all_nodes != NULL) {
       _numa_interleave_memory(start, size, _numa_all_nodes);
     }
   }
diff --git a/src/os_cpu/bsd_x86/vm/thread_bsd_x86.cpp b/src/os_cpu/bsd_x86/vm/thread_bsd_x86.cpp
index 8d167adaf5abf674a5bdf5f644172a5e8433fd49..628895649529ff8793b8fca84df950d7ea738f57 100644
--- a/src/os_cpu/bsd_x86/vm/thread_bsd_x86.cpp
+++ b/src/os_cpu/bsd_x86/vm/thread_bsd_x86.cpp
@@ -44,7 +44,7 @@ bool JavaThread::pd_get_top_frame(frame* fr_addr, void* ucontext, bool isInJava)
 
   // If we have a last_Java_frame, then we should use it even if
   // isInJava == true.  It should be more reliable than ucontext info.
-  if (jt->has_last_Java_frame()) {
+  if (jt->has_last_Java_frame() && jt->frame_anchor()->walkable()) {
     *fr_addr = jt->pd_last_frame();
     return true;
   }
diff --git a/src/os_cpu/bsd_x86/vm/thread_bsd_x86.hpp b/src/os_cpu/bsd_x86/vm/thread_bsd_x86.hpp
index 1d7921c711438df676a0b3520e5d8b2103df1500..b3bc5b6bee66f33496e078d49b18c0b500b7a488 100644
--- a/src/os_cpu/bsd_x86/vm/thread_bsd_x86.hpp
+++ b/src/os_cpu/bsd_x86/vm/thread_bsd_x86.hpp
@@ -32,12 +32,8 @@
 
   frame pd_last_frame() {
     assert(has_last_Java_frame(), "must have last_Java_sp() when suspended");
-    if (_anchor.last_Java_pc() != NULL) {
-      return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp(), _anchor.last_Java_pc());
-    } else {
-      // This will pick up pc from sp
-      return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp());
-    }
+    assert(_anchor.last_Java_pc() != NULL, "not walkable");
+    return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp(), _anchor.last_Java_pc());
   }
 
  public:
diff --git a/src/os_cpu/linux_x86/vm/thread_linux_x86.cpp b/src/os_cpu/linux_x86/vm/thread_linux_x86.cpp
index 77affdd52e6768f47afd963daa3808c9dbedaff3..2995b3c81eef5cf69e6fa6492502a67333620013 100644
--- a/src/os_cpu/linux_x86/vm/thread_linux_x86.cpp
+++ b/src/os_cpu/linux_x86/vm/thread_linux_x86.cpp
@@ -45,7 +45,7 @@ bool JavaThread::pd_get_top_frame(frame* fr_addr, void* ucontext, bool isInJava)
 
   // If we have a last_Java_frame, then we should use it even if
   // isInJava == true.  It should be more reliable than ucontext info.
-  if (jt->has_last_Java_frame()) {
+  if (jt->has_last_Java_frame() && jt->frame_anchor()->walkable()) {
     *fr_addr = jt->pd_last_frame();
     return true;
   }
diff --git a/src/os_cpu/linux_x86/vm/thread_linux_x86.hpp b/src/os_cpu/linux_x86/vm/thread_linux_x86.hpp
index 75fb7df5787a10f6e2e75e594ee112e57dc9a40c..bd8cc5a5e51382d28af3cdad6aabe528d2e56a35 100644
--- a/src/os_cpu/linux_x86/vm/thread_linux_x86.hpp
+++ b/src/os_cpu/linux_x86/vm/thread_linux_x86.hpp
@@ -32,12 +32,8 @@
 
   frame pd_last_frame() {
     assert(has_last_Java_frame(), "must have last_Java_sp() when suspended");
-    if (_anchor.last_Java_pc() != NULL) {
-      return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp(), _anchor.last_Java_pc());
-    } else {
-      // This will pick up pc from sp
-      return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp());
-    }
+    assert(_anchor.last_Java_pc() != NULL, "not walkable");
+    return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp(), _anchor.last_Java_pc());
   }
 
  public:
diff --git a/src/os_cpu/solaris_sparc/vm/vm_version_solaris_sparc.cpp b/src/os_cpu/solaris_sparc/vm/vm_version_solaris_sparc.cpp
index 0e1b7b307b35e9681296f4870243a66a2d2f41d5..b09f418757135841f0b8c4187f3de70cf96f7a23 100644
--- a/src/os_cpu/solaris_sparc/vm/vm_version_solaris_sparc.cpp
+++ b/src/os_cpu/solaris_sparc/vm/vm_version_solaris_sparc.cpp
@@ -442,7 +442,7 @@ int VM_Version::platform_features(int features) {
   // is available to us as well
   Sysinfo cpu_info(SI_CPUBRAND);
   bool use_solaris_12_api = cpu_info.valid();
-  const char* impl;
+  const char* impl = "unknown";
   int impl_m = 0;
   if (use_solaris_12_api) {
     impl = cpu_info.value();
@@ -477,7 +477,7 @@ int VM_Version::platform_features(int features) {
       kstat_close(kc);
     }
   }
-  assert(impl_m != 0, err_msg("Unknown CPU implementation %s", impl));
+  assert(impl_m != 0, err_msg("Unrecognized CPU implementation %s", impl));
   features |= impl_m;
 
   bool is_sun4v = (features & sun4v_m) != 0;
diff --git a/src/os_cpu/solaris_x86/vm/thread_solaris_x86.cpp b/src/os_cpu/solaris_x86/vm/thread_solaris_x86.cpp
index 3007c6bee7944a732532b0f39b12448d030a893d..b7a8b11d280822515124302ef4016c4114008072 100644
--- a/src/os_cpu/solaris_x86/vm/thread_solaris_x86.cpp
+++ b/src/os_cpu/solaris_x86/vm/thread_solaris_x86.cpp
@@ -44,9 +44,8 @@ bool JavaThread::pd_get_top_frame(frame* fr_addr,
   assert(this->is_Java_thread(), "must be JavaThread");
   JavaThread* jt = (JavaThread *)this;
 
-  // last_Java_frame is always walkable and safe use it if we have it
-
-  if (jt->has_last_Java_frame()) {
+  // There is small window where last_Java_frame is not walkable or safe
+  if (jt->has_last_Java_frame() && jt->frame_anchor()->walkable()) {
     *fr_addr = jt->pd_last_frame();
     return true;
   }
diff --git a/src/os_cpu/solaris_x86/vm/thread_solaris_x86.hpp b/src/os_cpu/solaris_x86/vm/thread_solaris_x86.hpp
index 7589a81a83b9aec5be897119ab1a372d266033d6..8484912d764e1eaac77e9d7aed52d5cd734dc33c 100644
--- a/src/os_cpu/solaris_x86/vm/thread_solaris_x86.hpp
+++ b/src/os_cpu/solaris_x86/vm/thread_solaris_x86.hpp
@@ -30,12 +30,8 @@
 
   frame pd_last_frame() {
     assert(has_last_Java_frame(), "must have last_Java_sp() when suspended");
-    if (_anchor.last_Java_pc() != NULL) {
-      return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp(), _anchor.last_Java_pc());
-    } else {
-      // This will pick up pc from sp
-      return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp());
-    }
+    assert(_anchor.last_Java_pc() != NULL, "not walkable");
+    return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp(), _anchor.last_Java_pc());
   }
 
  public:
diff --git a/src/os_cpu/windows_x86/vm/thread_windows_x86.cpp b/src/os_cpu/windows_x86/vm/thread_windows_x86.cpp
index 7180fe1d9fd5c3187027cad418eef96bd813caf7..95f47a680158621dc3075ca157b583f29ffe37ad 100644
--- a/src/os_cpu/windows_x86/vm/thread_windows_x86.cpp
+++ b/src/os_cpu/windows_x86/vm/thread_windows_x86.cpp
@@ -47,7 +47,7 @@ bool JavaThread::pd_get_top_frame(frame* fr_addr, void* ucontext, bool isInJava)
 
   // If we have a last_Java_frame, then we should use it even if
   // isInJava == true.  It should be more reliable than CONTEXT info.
-  if (jt->has_last_Java_frame()) {
+  if (jt->has_last_Java_frame() && jt->frame_anchor()->walkable()) {
     *fr_addr = jt->pd_last_frame();
     return true;
   }
diff --git a/src/os_cpu/windows_x86/vm/thread_windows_x86.hpp b/src/os_cpu/windows_x86/vm/thread_windows_x86.hpp
index 65aac35109a50a6ae62f16c407d4b59209b75713..e3350e8a65e2e7883bc453073514e8cbe101779f 100644
--- a/src/os_cpu/windows_x86/vm/thread_windows_x86.hpp
+++ b/src/os_cpu/windows_x86/vm/thread_windows_x86.hpp
@@ -32,12 +32,8 @@
 
   frame pd_last_frame() {
     assert(has_last_Java_frame(), "must have last_Java_sp() when suspended");
-    if (_anchor.last_Java_pc() != NULL) {
-      return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp(), _anchor.last_Java_pc());
-    } else {
-      // This will pick up pc from sp
-      return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp());
-    }
+    assert(_anchor.last_Java_pc() != NULL, "not walkable");
+    return frame(_anchor.last_Java_sp(), _anchor.last_Java_fp(), _anchor.last_Java_pc());
   }
 
  public:
diff --git a/src/share/vm/c1/c1_GraphBuilder.cpp b/src/share/vm/c1/c1_GraphBuilder.cpp
index b6d4180c85df8dc2baa163731c8fd7ae134b37e2..32185c65ef695350c09968b9884d712e4e8eea62 100644
--- a/src/share/vm/c1/c1_GraphBuilder.cpp
+++ b/src/share/vm/c1/c1_GraphBuilder.cpp
@@ -1530,7 +1530,7 @@ void GraphBuilder::method_return(Value x) {
         ciMethod* caller = state()->scope()->method();
         ciMethodData* md = caller->method_data_or_null();
         ciProfileData* data = md->bci_to_data(invoke_bci);
-        if (data->is_CallTypeData() || data->is_VirtualCallTypeData()) {
+        if (data != NULL && (data->is_CallTypeData() || data->is_VirtualCallTypeData())) {
           bool has_return = data->is_CallTypeData() ? ((ciCallTypeData*)data)->has_return() : ((ciVirtualCallTypeData*)data)->has_return();
           // May not be true in case of an inlined call through a method handle intrinsic.
           if (has_return) {
@@ -1747,7 +1747,7 @@ Values* GraphBuilder::args_list_for_profiling(ciMethod* target, int& start, bool
   start = has_receiver ? 1 : 0;
   if (profile_arguments()) {
     ciProfileData* data = method()->method_data()->bci_to_data(bci());
-    if (data->is_CallTypeData() || data->is_VirtualCallTypeData()) {
+    if (data != NULL && (data->is_CallTypeData() || data->is_VirtualCallTypeData())) {
       n = data->is_CallTypeData() ? data->as_CallTypeData()->number_of_arguments() : data->as_VirtualCallTypeData()->number_of_arguments();
     }
   }
@@ -4465,7 +4465,7 @@ void GraphBuilder::profile_return_type(Value ret, ciMethod* callee, ciMethod* m,
   }
   ciMethodData* md = m->method_data_or_null();
   ciProfileData* data = md->bci_to_data(invoke_bci);
-  if (data->is_CallTypeData() || data->is_VirtualCallTypeData()) {
+  if (data != NULL && (data->is_CallTypeData() || data->is_VirtualCallTypeData())) {
     append(new ProfileReturnType(m , invoke_bci, callee, ret));
   }
 }
diff --git a/src/share/vm/c1/c1_LIRGenerator.cpp b/src/share/vm/c1/c1_LIRGenerator.cpp
index 7d23f8c1010713a3a5d631cc03e5e1c4de6b89a1..2a4373ba03da7fad2b4b8cce5762ce8764bb6014 100644
--- a/src/share/vm/c1/c1_LIRGenerator.cpp
+++ b/src/share/vm/c1/c1_LIRGenerator.cpp
@@ -3185,50 +3185,52 @@ void LIRGenerator::profile_arguments(ProfileCall* x) {
     int bci = x->bci_of_invoke();
     ciMethodData* md = x->method()->method_data_or_null();
     ciProfileData* data = md->bci_to_data(bci);
-    if ((data->is_CallTypeData() && data->as_CallTypeData()->has_arguments()) ||
-        (data->is_VirtualCallTypeData() && data->as_VirtualCallTypeData()->has_arguments())) {
-      ByteSize extra = data->is_CallTypeData() ? CallTypeData::args_data_offset() : VirtualCallTypeData::args_data_offset();
-      int base_offset = md->byte_offset_of_slot(data, extra);
-      LIR_Opr mdp = LIR_OprFact::illegalOpr;
-      ciTypeStackSlotEntries* args = data->is_CallTypeData() ? ((ciCallTypeData*)data)->args() : ((ciVirtualCallTypeData*)data)->args();
-
-      Bytecodes::Code bc = x->method()->java_code_at_bci(bci);
-      int start = 0;
-      int stop = data->is_CallTypeData() ? ((ciCallTypeData*)data)->number_of_arguments() : ((ciVirtualCallTypeData*)data)->number_of_arguments();
-      if (x->callee()->is_loaded() && x->callee()->is_static() && Bytecodes::has_receiver(bc)) {
-        // first argument is not profiled at call (method handle invoke)
-        assert(x->method()->raw_code_at_bci(bci) == Bytecodes::_invokehandle, "invokehandle expected");
-        start = 1;
-      }
-      ciSignature* callee_signature = x->callee()->signature();
-      // method handle call to virtual method
-      bool has_receiver = x->callee()->is_loaded() && !x->callee()->is_static() && !Bytecodes::has_receiver(bc);
-      ciSignatureStream callee_signature_stream(callee_signature, has_receiver ? x->callee()->holder() : NULL);
-
-      bool ignored_will_link;
-      ciSignature* signature_at_call = NULL;
-      x->method()->get_method_at_bci(bci, ignored_will_link, &signature_at_call);
-      ciSignatureStream signature_at_call_stream(signature_at_call);
-
-      // if called through method handle invoke, some arguments may have been popped
-      for (int i = 0; i < stop && i+start < x->nb_profiled_args(); i++) {
-        int off = in_bytes(TypeEntriesAtCall::argument_type_offset(i)) - in_bytes(TypeEntriesAtCall::args_data_offset());
-        ciKlass* exact = profile_type(md, base_offset, off,
-                                      args->type(i), x->profiled_arg_at(i+start), mdp,
-                                      !x->arg_needs_null_check(i+start),
-                                      signature_at_call_stream.next_klass(), callee_signature_stream.next_klass());
-        if (exact != NULL) {
-          md->set_argument_type(bci, i, exact);
+    if (data != NULL) {
+      if ((data->is_CallTypeData() && data->as_CallTypeData()->has_arguments()) ||
+          (data->is_VirtualCallTypeData() && data->as_VirtualCallTypeData()->has_arguments())) {
+        ByteSize extra = data->is_CallTypeData() ? CallTypeData::args_data_offset() : VirtualCallTypeData::args_data_offset();
+        int base_offset = md->byte_offset_of_slot(data, extra);
+        LIR_Opr mdp = LIR_OprFact::illegalOpr;
+        ciTypeStackSlotEntries* args = data->is_CallTypeData() ? ((ciCallTypeData*)data)->args() : ((ciVirtualCallTypeData*)data)->args();
+
+        Bytecodes::Code bc = x->method()->java_code_at_bci(bci);
+        int start = 0;
+        int stop = data->is_CallTypeData() ? ((ciCallTypeData*)data)->number_of_arguments() : ((ciVirtualCallTypeData*)data)->number_of_arguments();
+        if (x->callee()->is_loaded() && x->callee()->is_static() && Bytecodes::has_receiver(bc)) {
+          // first argument is not profiled at call (method handle invoke)
+          assert(x->method()->raw_code_at_bci(bci) == Bytecodes::_invokehandle, "invokehandle expected");
+          start = 1;
         }
-      }
-    } else {
+        ciSignature* callee_signature = x->callee()->signature();
+        // method handle call to virtual method
+        bool has_receiver = x->callee()->is_loaded() && !x->callee()->is_static() && !Bytecodes::has_receiver(bc);
+        ciSignatureStream callee_signature_stream(callee_signature, has_receiver ? x->callee()->holder() : NULL);
+
+        bool ignored_will_link;
+        ciSignature* signature_at_call = NULL;
+        x->method()->get_method_at_bci(bci, ignored_will_link, &signature_at_call);
+        ciSignatureStream signature_at_call_stream(signature_at_call);
+
+        // if called through method handle invoke, some arguments may have been popped
+        for (int i = 0; i < stop && i+start < x->nb_profiled_args(); i++) {
+          int off = in_bytes(TypeEntriesAtCall::argument_type_offset(i)) - in_bytes(TypeEntriesAtCall::args_data_offset());
+          ciKlass* exact = profile_type(md, base_offset, off,
+              args->type(i), x->profiled_arg_at(i+start), mdp,
+              !x->arg_needs_null_check(i+start),
+              signature_at_call_stream.next_klass(), callee_signature_stream.next_klass());
+          if (exact != NULL) {
+            md->set_argument_type(bci, i, exact);
+          }
+        }
+      } else {
 #ifdef ASSERT
-      Bytecodes::Code code = x->method()->raw_code_at_bci(x->bci_of_invoke());
-      int n = x->nb_profiled_args();
-      assert(MethodData::profile_parameters() && (MethodData::profile_arguments_jsr292_only() ||
-                                                  (x->inlined() && ((code == Bytecodes::_invokedynamic && n <= 1) || (code == Bytecodes::_invokehandle && n <= 2)))),
-             "only at JSR292 bytecodes");
+        Bytecodes::Code code = x->method()->raw_code_at_bci(x->bci_of_invoke());
+        int n = x->nb_profiled_args();
+        assert(MethodData::profile_parameters() && (MethodData::profile_arguments_jsr292_only() ||
+            (x->inlined() && ((code == Bytecodes::_invokedynamic && n <= 1) || (code == Bytecodes::_invokehandle && n <= 2)))),
+            "only at JSR292 bytecodes");
 #endif
+      }
     }
   }
 }
@@ -3319,24 +3321,26 @@ void LIRGenerator::do_ProfileReturnType(ProfileReturnType* x) {
   int bci = x->bci_of_invoke();
   ciMethodData* md = x->method()->method_data_or_null();
   ciProfileData* data = md->bci_to_data(bci);
-  assert(data->is_CallTypeData() || data->is_VirtualCallTypeData(), "wrong profile data type");
-  ciReturnTypeEntry* ret = data->is_CallTypeData() ? ((ciCallTypeData*)data)->ret() : ((ciVirtualCallTypeData*)data)->ret();
-  LIR_Opr mdp = LIR_OprFact::illegalOpr;
-
-  bool ignored_will_link;
-  ciSignature* signature_at_call = NULL;
-  x->method()->get_method_at_bci(bci, ignored_will_link, &signature_at_call);
-
-  // The offset within the MDO of the entry to update may be too large
-  // to be used in load/store instructions on some platforms. So have
-  // profile_type() compute the address of the profile in a register.
-  ciKlass* exact = profile_type(md, md->byte_offset_of_slot(data, ret->type_offset()), 0,
-                                ret->type(), x->ret(), mdp,
-                                !x->needs_null_check(),
-                                signature_at_call->return_type()->as_klass(),
-                                x->callee()->signature()->return_type()->as_klass());
-  if (exact != NULL) {
-    md->set_return_type(bci, exact);
+  if (data != NULL) {
+    assert(data->is_CallTypeData() || data->is_VirtualCallTypeData(), "wrong profile data type");
+    ciReturnTypeEntry* ret = data->is_CallTypeData() ? ((ciCallTypeData*)data)->ret() : ((ciVirtualCallTypeData*)data)->ret();
+    LIR_Opr mdp = LIR_OprFact::illegalOpr;
+
+    bool ignored_will_link;
+    ciSignature* signature_at_call = NULL;
+    x->method()->get_method_at_bci(bci, ignored_will_link, &signature_at_call);
+
+    // The offset within the MDO of the entry to update may be too large
+    // to be used in load/store instructions on some platforms. So have
+    // profile_type() compute the address of the profile in a register.
+    ciKlass* exact = profile_type(md, md->byte_offset_of_slot(data, ret->type_offset()), 0,
+        ret->type(), x->ret(), mdp,
+        !x->needs_null_check(),
+        signature_at_call->return_type()->as_klass(),
+        x->callee()->signature()->return_type()->as_klass());
+    if (exact != NULL) {
+      md->set_return_type(bci, exact);
+    }
   }
 }
 
diff --git a/src/share/vm/ci/ciMethodData.cpp b/src/share/vm/ci/ciMethodData.cpp
index c57ec150daa83b647ccc26f6ec12b71049a3a186..a2d10796dc9f442f6483ac45e7dc5a5e743c7ca0 100644
--- a/src/share/vm/ci/ciMethodData.cpp
+++ b/src/share/vm/ci/ciMethodData.cpp
@@ -391,11 +391,13 @@ void ciMethodData::set_argument_type(int bci, int i, ciKlass* k) {
   MethodData* mdo = get_MethodData();
   if (mdo != NULL) {
     ProfileData* data = mdo->bci_to_data(bci);
-    if (data->is_CallTypeData()) {
-      data->as_CallTypeData()->set_argument_type(i, k->get_Klass());
-    } else {
-      assert(data->is_VirtualCallTypeData(), "no arguments!");
-      data->as_VirtualCallTypeData()->set_argument_type(i, k->get_Klass());
+    if (data != NULL) {
+      if (data->is_CallTypeData()) {
+        data->as_CallTypeData()->set_argument_type(i, k->get_Klass());
+      } else {
+        assert(data->is_VirtualCallTypeData(), "no arguments!");
+        data->as_VirtualCallTypeData()->set_argument_type(i, k->get_Klass());
+      }
     }
   }
 }
@@ -413,11 +415,13 @@ void ciMethodData::set_return_type(int bci, ciKlass* k) {
   MethodData* mdo = get_MethodData();
   if (mdo != NULL) {
     ProfileData* data = mdo->bci_to_data(bci);
-    if (data->is_CallTypeData()) {
-      data->as_CallTypeData()->set_return_type(k->get_Klass());
-    } else {
-      assert(data->is_VirtualCallTypeData(), "no arguments!");
-      data->as_VirtualCallTypeData()->set_return_type(k->get_Klass());
+    if (data != NULL) {
+      if (data->is_CallTypeData()) {
+        data->as_CallTypeData()->set_return_type(k->get_Klass());
+      } else {
+        assert(data->is_VirtualCallTypeData(), "no arguments!");
+        data->as_VirtualCallTypeData()->set_return_type(k->get_Klass());
+      }
     }
   }
 }
diff --git a/src/share/vm/opto/memnode.cpp b/src/share/vm/opto/memnode.cpp
index 2cfbae65be8a621a61f5b10c6d836495616ca2a0..1e76cbc797bf950321a9bc96b76f6e1d0288c776 100644
--- a/src/share/vm/opto/memnode.cpp
+++ b/src/share/vm/opto/memnode.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2016, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2017, 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
@@ -55,6 +55,15 @@ const TypePtr *MemNode::adr_type() const {
   return calculate_adr_type(adr->bottom_type(), cross_check);
 }
 
+bool MemNode::check_if_adr_maybe_raw(Node* adr) {
+  if (adr != NULL) {
+    if (adr->bottom_type()->base() == Type::RawPtr || adr->bottom_type()->base() == Type::AnyPtr) {
+      return true;
+    }
+  }
+  return false;
+}
+
 #ifndef PRODUCT
 void MemNode::dump_spec(outputStream *st) const {
   if (in(Address) == NULL)  return; // node is dead
@@ -503,6 +512,7 @@ Node* MemNode::find_previous_store(PhaseTransform* phase) {
   if (offset == Type::OffsetBot)
     return NULL;            // cannot unalias unless there are precise offsets
 
+  const bool adr_maybe_raw = check_if_adr_maybe_raw(adr);
   const TypeOopPtr *addr_t = adr->bottom_type()->isa_oopptr();
 
   intptr_t size_in_bytes = memory_size();
@@ -519,6 +529,13 @@ Node* MemNode::find_previous_store(PhaseTransform* phase) {
       Node* st_base = AddPNode::Ideal_base_and_offset(st_adr, phase, st_offset);
       if (st_base == NULL)
         break;              // inscrutable pointer
+
+      // For raw accesses it's not enough to prove that constant offsets don't intersect.
+      // We need the bases to be the equal in order for the offset check to make sense.
+      if ((adr_maybe_raw || check_if_adr_maybe_raw(st_adr)) && st_base != base) {
+        break;
+      }
+
       if (st_offset != offset && st_offset != Type::OffsetBot) {
         const int MAX_STORE = BytesPerLong;
         if (st_offset >= offset + size_in_bytes ||
diff --git a/src/share/vm/opto/memnode.hpp b/src/share/vm/opto/memnode.hpp
index 32d8af8638375dc7e7f38b0f08ca88fd9dafd0c6..e465833145b0139f51ef90ede83e1b4967654485 100644
--- a/src/share/vm/opto/memnode.hpp
+++ b/src/share/vm/opto/memnode.hpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2017, 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
@@ -75,6 +75,8 @@ protected:
     debug_only(_adr_type=at; adr_type();)
   }
 
+  static bool check_if_adr_maybe_raw(Node* adr);
+
 public:
   // Helpers for the optimizer.  Documented in memnode.cpp.
   static bool detect_ptr_independence(Node* p1, AllocateNode* a1,
diff --git a/src/share/vm/opto/parse2.cpp b/src/share/vm/opto/parse2.cpp
index 8bdd5b54bf889192d61ddd351ab59660f2d7de35..c4ce0e07d425469216838f6a2d8923f5fa9a6cbd 100644
--- a/src/share/vm/opto/parse2.cpp
+++ b/src/share/vm/opto/parse2.cpp
@@ -812,6 +812,9 @@ float Parse::dynamic_branch_prediction(float &cnt, BoolTest::mask btest, Node* t
     ciMethodData* methodData = method()->method_data();
     if (!methodData->is_mature())  return PROB_UNKNOWN;
     ciProfileData* data = methodData->bci_to_data(bci());
+    if (data == NULL) {
+      return PROB_UNKNOWN;
+    }
     if (!data->is_JumpData())  return PROB_UNKNOWN;
 
     // get taken and not taken values
@@ -903,8 +906,8 @@ float Parse::branch_prediction(float& cnt,
         // of the OSR-ed method, and we want to deopt to gather more stats.
         // If you have ANY counts, then this loop is simply 'cold' relative
         // to the OSR loop.
-        if (data->as_BranchData()->taken() +
-            data->as_BranchData()->not_taken() == 0 ) {
+        if (data == NULL ||
+            (data->as_BranchData()->taken() +  data->as_BranchData()->not_taken() == 0)) {
           // This is the only way to return PROB_UNKNOWN:
           return PROB_UNKNOWN;
         }
diff --git a/src/share/vm/services/lowMemoryDetector.cpp b/src/share/vm/services/lowMemoryDetector.cpp
index b68be23dfa4b91151198f430744ce6eb8447b37d..10bf5bcd76a35ced9fb6cec5a9fe886d6f457140 100644
--- a/src/share/vm/services/lowMemoryDetector.cpp
+++ b/src/share/vm/services/lowMemoryDetector.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2003, 2017, 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
@@ -298,19 +298,41 @@ void SensorInfo::trigger(int count, TRAPS) {
     Klass* k = Management::sun_management_Sensor_klass(CHECK);
     instanceKlassHandle sensorKlass (THREAD, k);
     Handle sensor_h(THREAD, _sensor_obj);
-    Handle usage_h = MemoryService::create_MemoryUsage_obj(_usage, CHECK);
+
+    Symbol* trigger_method_signature;
 
     JavaValue result(T_VOID);
     JavaCallArguments args(sensor_h);
     args.push_int((int) count);
-    args.push_oop(usage_h);
+
+    Handle usage_h = MemoryService::create_MemoryUsage_obj(_usage, THREAD);
+    // Call Sensor::trigger(int, MemoryUsage) to send notification to listeners.
+    // When OOME occurs and fails to allocate MemoryUsage object, call
+    // Sensor::trigger(int) instead.  The pending request will be processed
+    // but no notification will be sent.
+    if (HAS_PENDING_EXCEPTION) {
+       assert((PENDING_EXCEPTION->is_a(SystemDictionary::OutOfMemoryError_klass())), "we expect only an OOME here");
+       CLEAR_PENDING_EXCEPTION;
+       trigger_method_signature = vmSymbols::int_void_signature();
+    } else {
+       trigger_method_signature = vmSymbols::trigger_method_signature();
+       args.push_oop(usage_h);
+    }
 
     JavaCalls::call_virtual(&result,
                             sensorKlass,
                             vmSymbols::trigger_name(),
-                            vmSymbols::trigger_method_signature(),
+                            trigger_method_signature,
                             &args,
-                            CHECK);
+                            THREAD);
+
+    if (HAS_PENDING_EXCEPTION) {
+       // We just clear the OOM pending exception that we might have encountered
+       // in Java's tiggerAction(), and continue with updating the counters since
+       // the Java counters have been updated too.
+       assert((PENDING_EXCEPTION->is_a(SystemDictionary::OutOfMemoryError_klass())), "we expect only an OOME here");
+       CLEAR_PENDING_EXCEPTION;
+     }
   }
 
   {
diff --git a/test/compiler/unsafe/TestRawAliasing.java b/test/compiler/unsafe/TestRawAliasing.java
new file mode 100644
index 0000000000000000000000000000000000000000..515bfe1e23ee025fc6676f88c710901dc23cca57
--- /dev/null
+++ b/test/compiler/unsafe/TestRawAliasing.java
@@ -0,0 +1,70 @@
+/*
+ * Copyright (c) 2017, 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
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ * @test
+ * @bug 8178047
+ * @run main/othervm -XX:CompileCommand=exclude,*.main -XX:-TieredCompilation -XX:-BackgroundCompilation compiler.unsafe.TestRawAliasing
+ */
+
+package compiler.unsafe;
+
+import java.lang.reflect.Field;
+
+public class TestRawAliasing {
+    static private final sun.misc.Unsafe UNSAFE;
+    static {
+        try {
+            Field f = sun.misc.Unsafe.class.getDeclaredField("theUnsafe");
+            f.setAccessible(true);
+            UNSAFE = (sun.misc.Unsafe) f.get(null);
+        } catch (Exception e) {
+            throw new RuntimeException("Unable to get Unsafe instance.", e);
+        }
+    }
+
+    static private final int OFFSET_X = 50;
+    static private final int OFFSET_Y = 100;
+
+    private static int test(long base_plus_offset_x, long base_plus_offset_y, int magic_value) {
+        // write 0 to a location
+        UNSAFE.putByte(base_plus_offset_x - OFFSET_X, (byte)0);
+        // write unfoldable value to really the same location with another base
+        UNSAFE.putByte(base_plus_offset_y - OFFSET_Y, (byte)magic_value);
+        // read the value back, should be equal to "unfoldable_value"
+        return UNSAFE.getByte(base_plus_offset_x - OFFSET_X);
+    }
+
+    private static final int OFF_HEAP_AREA_SIZE = 128;
+    private static final byte MAGIC = 123;
+
+    // main is excluded from compilation since we don't want the test method to inline and make base values fold
+    public static void main(String... args) {
+        long base = UNSAFE.allocateMemory(OFF_HEAP_AREA_SIZE);
+        for (int i = 0; i < 100_000; i++) {
+            if (test(base + OFFSET_X, base + OFFSET_Y, MAGIC) != MAGIC) {
+                throw new RuntimeException("Unexpected magic value");
+            }
+        }
+    }
+}