7009361: JSR 292 Invalid value on stack on solaris-sparc with -Xcomp

Reviewed-by: kvn, twisti

18 files changed, 387 insertions, 172 deletions

diff --git a/src/cpu/sparc/vm/frame_sparc.cpp b/src/cpu/sparc/vm/frame_sparc.cpp
index e6fa47a5b..d9dd1700c 100644
--- a/src/cpu/sparc/vm/frame_sparc.cpp
+++ b/src/cpu/sparc/vm/frame_sparc.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2011, 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
@@ -806,3 +806,34 @@ intptr_t* frame::interpreter_frame_tos_at(jint offset) const {
   int index = (Interpreter::expr_offset_in_bytes(offset)/wordSize) - 1;
   return &interpreter_frame_tos_address()[index];
 }
+
+
+#ifdef ASSERT
+
+#define DESCRIBE_FP_OFFSET(name) \
+  values.describe(-1, fp() + frame::name##_offset, #name)
+
+void frame::describe_pd(FrameValues& values, int frame_no) {
+  for (int w = 0; w < frame::register_save_words; w++) {
+    values.describe(frame_no, sp() + w, err_msg("register save area word %d", w), 1);
+  }
+
+  if (is_interpreted_frame()) {
+    DESCRIBE_FP_OFFSET(interpreter_frame_d_scratch_fp);
+    DESCRIBE_FP_OFFSET(interpreter_frame_l_scratch_fp);
+    DESCRIBE_FP_OFFSET(interpreter_frame_padding);
+    DESCRIBE_FP_OFFSET(interpreter_frame_oop_temp);
+  }
+
+  if (!is_compiled_frame()) {
+    if (frame::callee_aggregate_return_pointer_words != 0) {
+      values.describe(frame_no, sp() + frame::callee_aggregate_return_pointer_sp_offset, "callee_aggregate_return_pointer_word");
+    }
+    for (int w = 0; w < frame::callee_register_argument_save_area_words; w++) {
+      values.describe(frame_no, sp() + frame::callee_register_argument_save_area_sp_offset + w,
+                      err_msg("callee_register_argument_save_area_words %d", w));
+    }
+  }
+}
+
+#endif
diff --git a/src/cpu/sparc/vm/methodHandles_sparc.cpp b/src/cpu/sparc/vm/methodHandles_sparc.cpp
index ac9e39fb6..0533fdab1 100644
--- a/src/cpu/sparc/vm/methodHandles_sparc.cpp
+++ b/src/cpu/sparc/vm/methodHandles_sparc.cpp
@@ -350,8 +350,9 @@ void MethodHandles::remove_arg_slots(MacroAssembler* _masm,
 #ifndef PRODUCT
 extern "C" void print_method_handle(oop mh);
 void trace_method_handle_stub(const char* adaptername,
-                              oopDesc* mh) {
-  printf("MH %s mh="INTPTR_FORMAT"\n", adaptername, (intptr_t) mh);
+                              oopDesc* mh,
+                              intptr_t* saved_sp) {
+  tty->print_cr("MH %s mh="INTPTR_FORMAT " saved_sp=" INTPTR_FORMAT, adaptername, (intptr_t) mh, saved_sp);
   print_method_handle(mh);
 }
 void MethodHandles::trace_method_handle(MacroAssembler* _masm, const char* adaptername) {
@@ -361,6 +362,7 @@ void MethodHandles::trace_method_handle(MacroAssembler* _masm, const char* adapt
   __ save_frame(16);
   __ set((intptr_t) adaptername, O0);
   __ mov(G3_method_handle, O1);
+  __ mov(I5_savedSP, O2);
   __ mov(G3_method_handle, L3);
   __ mov(Gargs, L4);
   __ mov(G5_method_type, L5);
diff --git a/src/cpu/sparc/vm/templateInterpreter_sparc.cpp b/src/cpu/sparc/vm/templateInterpreter_sparc.cpp
index 109eacfc4..ac855dad8 100644
--- a/src/cpu/sparc/vm/templateInterpreter_sparc.cpp
+++ b/src/cpu/sparc/vm/templateInterpreter_sparc.cpp
@@ -1698,35 +1698,21 @@ int AbstractInterpreter::layout_activation(methodOop method,
                      popframe_extra_args;
 
     int local_words = method->max_locals() * Interpreter::stackElementWords;
-    int parm_words  = method->size_of_parameters() * Interpreter::stackElementWords;
-    NEEDS_CLEANUP;
     intptr_t* locals;
-    if (caller->is_interpreted_frame()) {
-      // Can force the locals area to end up properly overlapping the top of the expression stack.
-      intptr_t* Lesp_ptr = caller->interpreter_frame_tos_address() - 1;
-      // Note that this computation means we replace size_of_parameters() values from the caller
-      // interpreter frame's expression stack with our argument locals
-      locals = Lesp_ptr + parm_words;
-      int delta = local_words - parm_words;
-      int computed_sp_adjustment = (delta > 0) ? round_to(delta, WordsPerLong) : 0;
-      *interpreter_frame->register_addr(I5_savedSP)    = (intptr_t) (fp + computed_sp_adjustment) - STACK_BIAS;
+    if (caller->is_compiled_frame()) {
+      // Compiled frames do not allocate a varargs area so place them
+      // next to the register save area.
+      locals = fp + frame::register_save_words + local_words - 1;
+      // Caller wants his own SP back
+      int caller_frame_size = caller->cb()->frame_size();
+      *interpreter_frame->register_addr(I5_savedSP) = (intptr_t)(caller->fp() - caller_frame_size) - STACK_BIAS;
     } else {
-      assert(caller->is_compiled_frame() || caller->is_entry_frame(), "only possible cases");
-      // Don't have Lesp available; lay out locals block in the caller
-      // adjacent to the register window save area.
-      //
-      // Compiled frames do not allocate a varargs area which is why this if
-      // statement is needed.
-      //
-      if (caller->is_compiled_frame()) {
-        locals = fp + frame::register_save_words + local_words - 1;
-      } else {
-        locals = fp + frame::memory_parameter_word_sp_offset + local_words - 1;
-      }
-      if (!caller->is_entry_frame()) {
-        // Caller wants his own SP back
-        int caller_frame_size = caller->cb()->frame_size();
-        *interpreter_frame->register_addr(I5_savedSP) = (intptr_t)(caller->fp() - caller_frame_size) - STACK_BIAS;
+      assert(caller->is_interpreted_frame() || caller->is_entry_frame(), "only possible cases");
+      // The entry and interpreter frames are laid out like normal C
+      // frames so place the locals adjacent to the varargs area.
+      locals = fp + frame::memory_parameter_word_sp_offset + local_words - 1;
+      if (caller->is_interpreted_frame()) {
+        *interpreter_frame->register_addr(I5_savedSP)    = (intptr_t) (fp + rounded_cls) - STACK_BIAS;
       }
     }
     if (TraceDeoptimization) {
diff --git a/src/cpu/x86/vm/assembler_x86.cpp b/src/cpu/x86/vm/assembler_x86.cpp
index 67e856d27..74e913019 100644
--- a/src/cpu/x86/vm/assembler_x86.cpp
+++ b/src/cpu/x86/vm/assembler_x86.cpp
@@ -6039,6 +6039,43 @@ void MacroAssembler::call_VM_leaf(address entry_point, Register arg_0, Register
   call_VM_leaf(entry_point, 3);
 }
 
+void MacroAssembler::super_call_VM_leaf(address entry_point, Register arg_0) {
+  pass_arg0(this, arg_0);
+  MacroAssembler::call_VM_leaf_base(entry_point, 1);
+}
+
+void MacroAssembler::super_call_VM_leaf(address entry_point, Register arg_0, Register arg_1) {
+
+  LP64_ONLY(assert(arg_0 != c_rarg1, "smashed arg"));
+  pass_arg1(this, arg_1);
+  pass_arg0(this, arg_0);
+  MacroAssembler::call_VM_leaf_base(entry_point, 2);
+}
+
+void MacroAssembler::super_call_VM_leaf(address entry_point, Register arg_0, Register arg_1, Register arg_2) {
+  LP64_ONLY(assert(arg_0 != c_rarg2, "smashed arg"));
+  LP64_ONLY(assert(arg_1 != c_rarg2, "smashed arg"));
+  pass_arg2(this, arg_2);
+  LP64_ONLY(assert(arg_0 != c_rarg1, "smashed arg"));
+  pass_arg1(this, arg_1);
+  pass_arg0(this, arg_0);
+  MacroAssembler::call_VM_leaf_base(entry_point, 3);
+}
+
+void MacroAssembler::super_call_VM_leaf(address entry_point, Register arg_0, Register arg_1, Register arg_2, Register arg_3) {
+  LP64_ONLY(assert(arg_0 != c_rarg3, "smashed arg"));
+  LP64_ONLY(assert(arg_1 != c_rarg3, "smashed arg"));
+  LP64_ONLY(assert(arg_2 != c_rarg3, "smashed arg"));
+  pass_arg3(this, arg_3);
+  LP64_ONLY(assert(arg_0 != c_rarg2, "smashed arg"));
+  LP64_ONLY(assert(arg_1 != c_rarg2, "smashed arg"));
+  pass_arg2(this, arg_2);
+  LP64_ONLY(assert(arg_0 != c_rarg1, "smashed arg"));
+  pass_arg1(this, arg_1);
+  pass_arg0(this, arg_0);
+  MacroAssembler::call_VM_leaf_base(entry_point, 4);
+}
+
 void MacroAssembler::check_and_handle_earlyret(Register java_thread) {
 }
 
diff --git a/src/cpu/x86/vm/assembler_x86.hpp b/src/cpu/x86/vm/assembler_x86.hpp
index 4c6b32faf..34e8ed9c8 100644
--- a/src/cpu/x86/vm/assembler_x86.hpp
+++ b/src/cpu/x86/vm/assembler_x86.hpp
@@ -1655,6 +1655,14 @@ class MacroAssembler: public Assembler {
   void call_VM_leaf(address entry_point,
                     Register arg_1, Register arg_2, Register arg_3);
 
+  // These always tightly bind to MacroAssembler::call_VM_leaf_base
+  // bypassing the virtual implementation
+  void super_call_VM_leaf(address entry_point);
+  void super_call_VM_leaf(address entry_point, Register arg_1);
+  void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2);
+  void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2, Register arg_3);
+  void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2, Register arg_3, Register arg_4);
+
   // last Java Frame (fills frame anchor)
   void set_last_Java_frame(Register thread,
                            Register last_java_sp,
diff --git a/src/cpu/x86/vm/frame_x86.cpp b/src/cpu/x86/vm/frame_x86.cpp
index 8f1585fb9..8709a0e86 100644
--- a/src/cpu/x86/vm/frame_x86.cpp
+++ b/src/cpu/x86/vm/frame_x86.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2011, 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
@@ -669,3 +669,23 @@ intptr_t* frame::interpreter_frame_tos_at(jint offset) const {
   int index = (Interpreter::expr_offset_in_bytes(offset)/wordSize);
   return &interpreter_frame_tos_address()[index];
 }
+
+#ifdef ASSERT
+
+#define DESCRIBE_FP_OFFSET(name) \
+  values.describe(-1, fp() + frame::name##_offset, #name)
+
+void frame::describe_pd(FrameValues& values, int frame_no) {
+  if (is_interpreted_frame()) {
+    DESCRIBE_FP_OFFSET(interpreter_frame_sender_sp);
+    DESCRIBE_FP_OFFSET(interpreter_frame_last_sp);
+    DESCRIBE_FP_OFFSET(interpreter_frame_method);
+    DESCRIBE_FP_OFFSET(interpreter_frame_mdx);
+    DESCRIBE_FP_OFFSET(interpreter_frame_cache);
+    DESCRIBE_FP_OFFSET(interpreter_frame_locals);
+    DESCRIBE_FP_OFFSET(interpreter_frame_bcx);
+    DESCRIBE_FP_OFFSET(interpreter_frame_initial_sp);
+  }
+
+}
+#endif
diff --git a/src/cpu/x86/vm/interp_masm_x86_32.cpp b/src/cpu/x86/vm/interp_masm_x86_32.cpp
index e0d5cf03c..ce3571338 100644
--- a/src/cpu/x86/vm/interp_masm_x86_32.cpp
+++ b/src/cpu/x86/vm/interp_masm_x86_32.cpp
@@ -383,32 +383,6 @@ void InterpreterMacroAssembler::store_ptr(int n, Register val) {
   movptr(Address(rsp, Interpreter::expr_offset_in_bytes(n)), val);
 }
 
-void InterpreterMacroAssembler::super_call_VM_leaf(address entry_point) {
-  MacroAssembler::call_VM_leaf_base(entry_point, 0);
-}
-
-
-void InterpreterMacroAssembler::super_call_VM_leaf(address entry_point, Register arg_1) {
-  push(arg_1);
-  MacroAssembler::call_VM_leaf_base(entry_point, 1);
-}
-
-
-void InterpreterMacroAssembler::super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2) {
-  push(arg_2);
-  push(arg_1);
-  MacroAssembler::call_VM_leaf_base(entry_point, 2);
-}
-
-
-void InterpreterMacroAssembler::super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2, Register arg_3) {
-  push(arg_3);
-  push(arg_2);
-  push(arg_1);
-  MacroAssembler::call_VM_leaf_base(entry_point, 3);
-}
-
-
 void InterpreterMacroAssembler::prepare_to_jump_from_interpreted() {
   // set sender sp
   lea(rsi, Address(rsp, wordSize));
diff --git a/src/cpu/x86/vm/interp_masm_x86_32.hpp b/src/cpu/x86/vm/interp_masm_x86_32.hpp
index 404b4d572..d08f0fd59 100644
--- a/src/cpu/x86/vm/interp_masm_x86_32.hpp
+++ b/src/cpu/x86/vm/interp_masm_x86_32.hpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2011, 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
@@ -124,12 +124,6 @@ class InterpreterMacroAssembler: public MacroAssembler {
   void load_ptr(int n, Register val);
   void store_ptr(int n, Register val);
 
-  // Super call_VM calls - correspond to MacroAssembler::call_VM(_leaf) calls
-  void super_call_VM_leaf(address entry_point);
-  void super_call_VM_leaf(address entry_point, Register arg_1);
-  void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2);
-  void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2, Register arg_3);
-
   // Generate a subtype check: branch to ok_is_subtype if sub_klass is
   // a subtype of super_klass.  EAX holds the super_klass.  Blows ECX
   // and EDI.  Register sub_klass cannot be any of the above.
diff --git a/src/cpu/x86/vm/interp_masm_x86_64.cpp b/src/cpu/x86/vm/interp_masm_x86_64.cpp
index 428281d00..4b7884cec 100644
--- a/src/cpu/x86/vm/interp_masm_x86_64.cpp
+++ b/src/cpu/x86/vm/interp_masm_x86_64.cpp
@@ -381,56 +381,6 @@ void InterpreterMacroAssembler::store_ptr(int n, Register val) {
 }
 
 
-void InterpreterMacroAssembler::super_call_VM_leaf(address entry_point) {
-  MacroAssembler::call_VM_leaf_base(entry_point, 0);
-}
-
-
-void InterpreterMacroAssembler::super_call_VM_leaf(address entry_point,
-                                                   Register arg_1) {
-  if (c_rarg0 != arg_1) {
-    mov(c_rarg0, arg_1);
-  }
-  MacroAssembler::call_VM_leaf_base(entry_point, 1);
-}
-
-
-void InterpreterMacroAssembler::super_call_VM_leaf(address entry_point,
-                                                   Register arg_1,
-                                                   Register arg_2) {
-  assert(c_rarg0 != arg_2, "smashed argument");
-  assert(c_rarg1 != arg_1, "smashed argument");
-  if (c_rarg0 != arg_1) {
-    mov(c_rarg0, arg_1);
-  }
-  if (c_rarg1 != arg_2) {
-    mov(c_rarg1, arg_2);
-  }
-  MacroAssembler::call_VM_leaf_base(entry_point, 2);
-}
-
-void InterpreterMacroAssembler::super_call_VM_leaf(address entry_point,
-                                                   Register arg_1,
-                                                   Register arg_2,
-                                                   Register arg_3) {
-  assert(c_rarg0 != arg_2, "smashed argument");
-  assert(c_rarg0 != arg_3, "smashed argument");
-  assert(c_rarg1 != arg_1, "smashed argument");
-  assert(c_rarg1 != arg_3, "smashed argument");
-  assert(c_rarg2 != arg_1, "smashed argument");
-  assert(c_rarg2 != arg_2, "smashed argument");
-  if (c_rarg0 != arg_1) {
-    mov(c_rarg0, arg_1);
-  }
-  if (c_rarg1 != arg_2) {
-    mov(c_rarg1, arg_2);
-  }
-  if (c_rarg2 != arg_3) {
-    mov(c_rarg2, arg_3);
-  }
-  MacroAssembler::call_VM_leaf_base(entry_point, 3);
-}
-
 void InterpreterMacroAssembler::prepare_to_jump_from_interpreted() {
   // set sender sp
   lea(r13, Address(rsp, wordSize));
diff --git a/src/cpu/x86/vm/interp_masm_x86_64.hpp b/src/cpu/x86/vm/interp_masm_x86_64.hpp
index 62539a735..b203abc7d 100644
--- a/src/cpu/x86/vm/interp_masm_x86_64.hpp
+++ b/src/cpu/x86/vm/interp_masm_x86_64.hpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2003, 2011, 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
@@ -136,13 +136,6 @@ class InterpreterMacroAssembler: public MacroAssembler {
   void load_ptr(int n, Register val);
   void store_ptr(int n, Register val);
 
-  // Super call_VM calls - correspond to MacroAssembler::call_VM(_leaf) calls
-  void super_call_VM_leaf(address entry_point);
-  void super_call_VM_leaf(address entry_point, Register arg_1);
-  void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2);
-  void super_call_VM_leaf(address entry_point,
-                          Register arg_1, Register arg_2, Register arg_3);
-
   // Generate a subtype check: branch to ok_is_subtype if sub_klass is
   // a subtype of super_klass.
   void gen_subtype_check( Register sub_klass, Label &ok_is_subtype );
diff --git a/src/cpu/x86/vm/methodHandles_x86.cpp b/src/cpu/x86/vm/methodHandles_x86.cpp
index 704fb7d35..42a57c132 100644
--- a/src/cpu/x86/vm/methodHandles_x86.cpp
+++ b/src/cpu/x86/vm/methodHandles_x86.cpp
@@ -315,56 +315,38 @@ void MethodHandles::remove_arg_slots(MacroAssembler* _masm,
 #ifndef PRODUCT
 extern "C" void print_method_handle(oop mh);
 void trace_method_handle_stub(const char* adaptername,
-                              oop mh,
-                              intptr_t* saved_regs,
-                              intptr_t* entry_sp,
                               intptr_t* saved_sp,
-                              intptr_t* saved_bp) {
+                              oop mh,
+                              intptr_t* sp) {
   // called as a leaf from native code: do not block the JVM!
-  intptr_t* last_sp = (intptr_t*) saved_bp[frame::interpreter_frame_last_sp_offset];
-  intptr_t* base_sp = (intptr_t*) saved_bp[frame::interpreter_frame_monitor_block_top_offset];
-  printf("MH %s mh="INTPTR_FORMAT" sp=("INTPTR_FORMAT"+"INTX_FORMAT") stack_size="INTX_FORMAT" bp="INTPTR_FORMAT"\n",
-         adaptername, (intptr_t)mh, (intptr_t)entry_sp, (intptr_t)(saved_sp - entry_sp), (intptr_t)(base_sp - last_sp), (intptr_t)saved_bp);
-  if (last_sp != saved_sp && last_sp != NULL)
-    printf("*** last_sp="INTPTR_FORMAT"\n", (intptr_t)last_sp);
+  intptr_t* entry_sp = sp + LP64_ONLY(16) NOT_LP64(8);
+  tty->print_cr("MH %s mh="INTPTR_FORMAT" sp="INTPTR_FORMAT" saved_sp="INTPTR_FORMAT")",
+                adaptername, (intptr_t)mh, (intptr_t)entry_sp, saved_sp);
   if (Verbose) {
-    printf(" reg dump: ");
-    int saved_regs_count = (entry_sp-1) - saved_regs;
-    // 32 bit: rdi rsi rbp rsp; rbx rdx rcx (*) rax
-    int i;
-    for (i = 0; i <= saved_regs_count; i++) {
-      if (i > 0 && i % 4 == 0 && i != saved_regs_count)
-        printf("\n   + dump: ");
-      printf(" %d: "INTPTR_FORMAT, i, saved_regs[i]);
-    }
-    printf("\n");
-    int stack_dump_count = 16;
-    if (stack_dump_count < (int)(saved_bp + 2 - saved_sp))
-      stack_dump_count = (int)(saved_bp + 2 - saved_sp);
-    if (stack_dump_count > 64)  stack_dump_count = 48;
-    for (i = 0; i < stack_dump_count; i += 4) {
-      printf(" dump at SP[%d] "INTPTR_FORMAT": "INTPTR_FORMAT" "INTPTR_FORMAT" "INTPTR_FORMAT" "INTPTR_FORMAT"\n",
-             i, (intptr_t) &entry_sp[i+0], entry_sp[i+0], entry_sp[i+1], entry_sp[i+2], entry_sp[i+3]);
-    }
     print_method_handle(mh);
   }
 }
 void MethodHandles::trace_method_handle(MacroAssembler* _masm, const char* adaptername) {
   if (!TraceMethodHandles)  return;
   BLOCK_COMMENT("trace_method_handle {");
-  __ push(rax);
-  __ lea(rax, Address(rsp, wordSize*6)); // entry_sp
   __ pusha();
+#ifdef _LP64
+  // Pass arguments carefully since the registers overlap with the calling convention.
+  // rcx: method handle
+  // r13: saved sp
+  __ mov(c_rarg2, rcx); // mh
+  __ mov(c_rarg1, r13); // saved sp
+  __ mov(c_rarg3, rsp); // sp
+  __ movptr(c_rarg0, (intptr_t) adaptername);
+  __ super_call_VM_leaf(CAST_FROM_FN_PTR(address, trace_method_handle_stub), c_rarg0, c_rarg1, c_rarg2, c_rarg3);
+#else
   // arguments:
-  __ push(rbp);               // interpreter frame pointer
-  __ push(rsi);               // saved_sp
-  __ push(rax);               // entry_sp
-  __ push(rcx);               // mh
-  __ push(rcx);
-  __ movptr(Address(rsp, 0), (intptr_t) adaptername);
-  __ call_VM_leaf(CAST_FROM_FN_PTR(address, trace_method_handle_stub), 5);
+  // rcx: method handle
+  // rsi: saved sp
+  __ movptr(rbx, (intptr_t) adaptername);
+  __ super_call_VM_leaf(CAST_FROM_FN_PTR(address, trace_method_handle_stub), rbx, rsi, rcx, rsp);
+#endif
   __ popa();
-  __ pop(rax);
   BLOCK_COMMENT("} trace_method_handle");
 }
 #endif //PRODUCT
diff --git a/src/share/vm/opto/loopPredicate.cpp b/src/share/vm/opto/loopPredicate.cpp
index c5ac2c682..d436a572c 100644
--- a/src/share/vm/opto/loopPredicate.cpp
+++ b/src/share/vm/opto/loopPredicate.cpp
@@ -700,32 +700,35 @@ BoolNode* PhaseIdealLoop::rc_predicate(Node* ctrl,
                                        int scale, Node* offset,
                                        Node* init, Node* limit, Node* stride,
                                        Node* range, bool upper) {
-  DEBUG_ONLY(ttyLocker ttyl);
-  if (TraceLoopPredicate) tty->print("rc_predicate ");
+  stringStream* predString = NULL;
+  if (TraceLoopPredicate) {
+    predString = new stringStream();
+    predString->print("rc_predicate ");
+  }
 
   Node* max_idx_expr  = init;
   int stride_con = stride->get_int();
   if ((stride_con > 0) == (scale > 0) == upper) {
     max_idx_expr = new (C, 3) SubINode(limit, stride);
     register_new_node(max_idx_expr, ctrl);
-    if (TraceLoopPredicate) tty->print("(limit - stride) ");
+    if (TraceLoopPredicate) predString->print("(limit - stride) ");
   } else {
-    if (TraceLoopPredicate) tty->print("init ");
+    if (TraceLoopPredicate) predString->print("init ");
   }
 
   if (scale != 1) {
     ConNode* con_scale = _igvn.intcon(scale);
     max_idx_expr = new (C, 3) MulINode(max_idx_expr, con_scale);
     register_new_node(max_idx_expr, ctrl);
-    if (TraceLoopPredicate) tty->print("* %d ", scale);
+    if (TraceLoopPredicate) predString->print("* %d ", scale);
   }
 
   if (offset && (!offset->is_Con() || offset->get_int() != 0)){
     max_idx_expr = new (C, 3) AddINode(max_idx_expr, offset);
     register_new_node(max_idx_expr, ctrl);
     if (TraceLoopPredicate)
-      if (offset->is_Con()) tty->print("+ %d ", offset->get_int());
-      else tty->print("+ offset ");
+      if (offset->is_Con()) predString->print("+ %d ", offset->get_int());
+      else predString->print("+ offset ");
   }
 
   CmpUNode* cmp = new (C, 3) CmpUNode(max_idx_expr, range);
@@ -733,7 +736,10 @@ BoolNode* PhaseIdealLoop::rc_predicate(Node* ctrl,
   BoolNode* bol = new (C, 2) BoolNode(cmp, BoolTest::lt);
   register_new_node(bol, ctrl);
 
-  if (TraceLoopPredicate) tty->print_cr("<u range");
+  if (TraceLoopPredicate) {
+    predString->print_cr("<u range");
+    tty->print(predString->as_string());
+  }
   return bol;
 }
 
@@ -957,4 +963,3 @@ bool IdealLoopTree::loop_predication( PhaseIdealLoop *phase) {
 
   return hoisted;
 }
-
diff --git a/src/share/vm/runtime/deoptimization.cpp b/src/share/vm/runtime/deoptimization.cpp
index 3e03486d3..10ebd8364 100644
--- a/src/share/vm/runtime/deoptimization.cpp
+++ b/src/share/vm/runtime/deoptimization.cpp
@@ -189,6 +189,10 @@ Deoptimization::UnrollBlock* Deoptimization::fetch_unroll_info_helper(JavaThread
   assert(thread->deopt_nmethod() == NULL, "Pending deopt!");
   thread->set_deopt_nmethod(deoptee.cb()->as_nmethod_or_null());
 
+  if (VerifyStack) {
+    thread->validate_frame_layout();
+  }
+
   // Create a growable array of VFrames where each VFrame represents an inlined
   // Java frame.  This storage is allocated with the usual system arena.
   assert(deoptee.is_compiled_frame(), "Wrong frame type");
@@ -421,6 +425,21 @@ Deoptimization::UnrollBlock* Deoptimization::fetch_unroll_info_helper(JavaThread
   frame deopt_sender = stub_frame.sender(&dummy_map); // First is the deoptee frame
   deopt_sender = deopt_sender.sender(&dummy_map);     // Now deoptee caller
 
+  // It's possible that the number of paramters at the call site is
+  // different than number of arguments in the callee when method
+  // handles are used.  If the caller is interpreted get the real
+  // value so that the proper amount of space can be added to it's
+  // frame.
+  int sender_callee_parameters = callee_parameters;
+  if (deopt_sender.is_interpreted_frame()) {
+    methodHandle method = deopt_sender.interpreter_frame_method();
+    Bytecode_invoke cur = Bytecode_invoke_check(method,
+                                                deopt_sender.interpreter_frame_bci());
+    Symbol* signature = method->constants()->signature_ref_at(cur.index());
+    ArgumentSizeComputer asc(signature);
+    sender_callee_parameters = asc.size() + cur.has_receiver() ? 1 : 0;
+  }
+
   // Compute the amount the oldest interpreter frame will have to adjust
   // its caller's stack by. If the caller is a compiled frame then
   // we pretend that the callee has no parameters so that the
@@ -435,14 +454,13 @@ Deoptimization::UnrollBlock* Deoptimization::fetch_unroll_info_helper(JavaThread
 
   if (deopt_sender.is_compiled_frame()) {
     caller_adjustment = last_frame_adjust(0, callee_locals);
-  } else if (callee_locals > callee_parameters) {
+  } else if (callee_locals > sender_callee_parameters) {
     // The caller frame may need extending to accommodate
     // non-parameter locals of the first unpacked interpreted frame.
     // Compute that adjustment.
-    caller_adjustment = last_frame_adjust(callee_parameters, callee_locals);
+    caller_adjustment = last_frame_adjust(sender_callee_parameters, callee_locals);
   }
 
-
   // If the sender is deoptimized the we must retrieve the address of the handler
   // since the frame will "magically" show the original pc before the deopt
   // and we'd undo the deopt.
@@ -569,6 +587,8 @@ JRT_LEAF(BasicType, Deoptimization::unpack_frames(JavaThread* thread, int exec_m
   if (VerifyStack) {
     ResourceMark res_mark;
 
+    thread->validate_frame_layout();
+
     // Verify that the just-unpacked frames match the interpreter's
     // notions of expression stack and locals
     vframeArray* cur_array = thread->vframe_array_last();
diff --git a/src/share/vm/runtime/frame.cpp b/src/share/vm/runtime/frame.cpp
index 5c205501a..b5352fc00 100644
--- a/src/share/vm/runtime/frame.cpp
+++ b/src/share/vm/runtime/frame.cpp
@@ -1308,6 +1308,72 @@ void frame::interpreter_frame_verify_monitor(BasicObjectLock* value) const {
   guarantee((current - low_mark) % monitor_size  ==  0         , "Misaligned bottom of BasicObjectLock*");
   guarantee( current >= low_mark                               , "Current BasicObjectLock* below than low_mark");
 }
+
+
+void frame::describe(FrameValues& values, int frame_no) {
+  if (is_entry_frame() || is_compiled_frame() || is_interpreted_frame() || is_native_frame()) {
+    // Label values common to most frames
+    values.describe(-1, unextended_sp(), err_msg("unextended_sp for #%d", frame_no));
+    values.describe(-1, sp(), err_msg("sp for #%d", frame_no));
+    values.describe(-1, fp(), err_msg("fp for #%d", frame_no));
+  }
+  if (is_interpreted_frame()) {
+    methodOop m = interpreter_frame_method();
+    int bci = interpreter_frame_bci();
+
+    // Label the method and current bci
+    values.describe(-1, MAX2(sp(), fp()),
+                    FormatBuffer<1024>("#%d method %s @ %d", frame_no, m->name_and_sig_as_C_string(), bci), 2);
+    values.describe(-1, MAX2(sp(), fp()),
+                    err_msg("- %d locals %d max stack", m->max_locals(), m->max_stack()), 1);
+    if (m->max_locals() > 0) {
+      intptr_t* l0 = interpreter_frame_local_at(0);
+      intptr_t* ln = interpreter_frame_local_at(m->max_locals() - 1);
+      values.describe(-1, MAX2(l0, ln), err_msg("locals for #%d", frame_no), 1);
+      // Report each local and mark as owned by this frame
+      for (int l = 0; l < m->max_locals(); l++) {
+        intptr_t* l0 = interpreter_frame_local_at(l);
+        values.describe(frame_no, l0, err_msg("local %d", l));
+      }
+    }
+
+    // Compute the actual expression stack size
+    InterpreterOopMap mask;
+    OopMapCache::compute_one_oop_map(m, bci, &mask);
+    intptr_t* tos = NULL;
+    // Report each stack element and mark as owned by this frame
+    for (int e = 0; e < mask.expression_stack_size(); e++) {
+      tos = MAX2(tos, interpreter_frame_expression_stack_at(e));
+      values.describe(frame_no, interpreter_frame_expression_stack_at(e),
+                      err_msg("stack %d", e));
+    }
+    if (tos != NULL) {
+      values.describe(-1, tos, err_msg("expression stack for #%d", frame_no), 1);
+    }
+    if (interpreter_frame_monitor_begin() != interpreter_frame_monitor_end()) {
+      values.describe(frame_no, (intptr_t*)interpreter_frame_monitor_begin(), "monitors begin");
+      values.describe(frame_no, (intptr_t*)interpreter_frame_monitor_end(), "monitors end");
+    }
+  } else if (is_entry_frame()) {
+    // For now just label the frame
+    values.describe(-1, MAX2(sp(), fp()), err_msg("#%d entry frame", frame_no), 2);
+  } else if (is_compiled_frame()) {
+    // For now just label the frame
+    nmethod* nm = cb()->as_nmethod_or_null();
+    values.describe(-1, MAX2(sp(), fp()),
+                    FormatBuffer<1024>("#%d nmethod " INTPTR_FORMAT " for method %s%s", frame_no,
+                                       nm, nm->method()->name_and_sig_as_C_string(),
+                                       is_deoptimized_frame() ? " (deoptimized" : ""), 2);
+  } else if (is_native_frame()) {
+    // For now just label the frame
+    nmethod* nm = cb()->as_nmethod_or_null();
+    values.describe(-1, MAX2(sp(), fp()),
+                    FormatBuffer<1024>("#%d nmethod " INTPTR_FORMAT " for native method %s", frame_no,
+                                       nm, nm->method()->name_and_sig_as_C_string()), 2);
+  }
+  describe_pd(values, frame_no);
+}
+
 #endif
 
 
@@ -1319,3 +1385,71 @@ StackFrameStream::StackFrameStream(JavaThread *thread, bool update) : _reg_map(t
   _fr = thread->last_frame();
   _is_done = false;
 }
+
+
+#ifdef ASSERT
+
+void FrameValues::describe(int owner, intptr_t* location, const char* description, int priority) {
+  FrameValue fv;
+  fv.location = location;
+  fv.owner = owner;
+  fv.priority = priority;
+  fv.description = NEW_RESOURCE_ARRAY(char, strlen(description) + 1);
+  strcpy(fv.description, description);
+  _values.append(fv);
+}
+
+
+bool FrameValues::validate() {
+  _values.sort(compare);
+  bool error = false;
+  FrameValue prev;
+  prev.owner = -1;
+  for (int i = _values.length() - 1; i >= 0; i--) {
+    FrameValue fv = _values.at(i);
+    if (fv.owner == -1) continue;
+    if (prev.owner == -1) {
+      prev = fv;
+      continue;
+    }
+    if (prev.location == fv.location) {
+      if (fv.owner != prev.owner) {
+        tty->print_cr("overlapping storage");
+        tty->print_cr(" " INTPTR_FORMAT ": " INTPTR_FORMAT " %s", prev.location, *prev.location, prev.description);
+        tty->print_cr(" " INTPTR_FORMAT ": " INTPTR_FORMAT " %s", fv.location, *fv.location, fv.description);
+        error = true;
+      }
+    } else {
+      prev = fv;
+    }
+  }
+  return error;
+}
+
+
+void FrameValues::print() {
+  _values.sort(compare);
+  intptr_t* v0 = _values.at(0).location;
+  intptr_t* v1 = _values.at(_values.length() - 1).location;
+  intptr_t* min = MIN2(v0, v1);
+  intptr_t* max = MAX2(v0, v1);
+  intptr_t* cur = max;
+  intptr_t* last = NULL;
+  for (int i = _values.length() - 1; i >= 0; i--) {
+    FrameValue fv = _values.at(i);
+    while (cur > fv.location) {
+      tty->print_cr(" " INTPTR_FORMAT ": " INTPTR_FORMAT, cur, *cur);
+      cur--;
+    }
+    if (last == fv.location) {
+      const char* spacer = "          " LP64_ONLY("        ");
+      tty->print_cr(" %s  %s %s", spacer, spacer, fv.description);
+    } else {
+      tty->print_cr(" " INTPTR_FORMAT ": " INTPTR_FORMAT " %s", fv.location, *fv.location, fv.description);
+      last = fv.location;
+      cur--;
+    }
+  }
+}
+
+#endif
diff --git a/src/share/vm/runtime/frame.hpp b/src/share/vm/runtime/frame.hpp
index a48c2af37..66d17a9f7 100644
--- a/src/share/vm/runtime/frame.hpp
+++ b/src/share/vm/runtime/frame.hpp
@@ -60,6 +60,7 @@
 typedef class BytecodeInterpreter* interpreterState;
 
 class CodeBlob;
+class FrameValues;
 class vframeArray;
 
 
@@ -381,6 +382,8 @@ class frame VALUE_OBJ_CLASS_SPEC {
  private:
   const char* print_name() const;
 
+  void describe_pd(FrameValues& values, int frame_no);
+
  public:
   void print_value() const { print_value_on(tty,NULL); }
   void print_value_on(outputStream* st, JavaThread *thread) const;
@@ -388,6 +391,9 @@ class frame VALUE_OBJ_CLASS_SPEC {
   void interpreter_frame_print_on(outputStream* st) const;
   void print_on_error(outputStream* st, char* buf, int buflen, bool verbose = false) const;
 
+  // Add annotated descriptions of memory locations belonging to this frame to values
+  void describe(FrameValues& values, int frame_no);
+
   // Conversion from an VMReg to physical stack location
   oop* oopmapreg_to_location(VMReg reg, const RegisterMap* regmap) const;
 
@@ -472,6 +478,41 @@ class frame VALUE_OBJ_CLASS_SPEC {
 
 };
 
+#ifdef ASSERT
+// A simple class to describe a location on the stack
+class FrameValue VALUE_OBJ_CLASS_SPEC {
+ public:
+  intptr_t* location;
+  char* description;
+  int owner;
+  int priority;
+};
+
+
+// A collection of described stack values that can print a symbolic
+// description of the stack memory.  Interpreter frame values can be
+// in the caller frames so all the values are collected first and then
+// sorted before being printed.
+class FrameValues {
+ private:
+  GrowableArray<FrameValue> _values;
+
+  static int compare(FrameValue* a, FrameValue* b) {
+    if (a->location == b->location) {
+      return a->priority - b->priority;
+    }
+    return a->location - b->location;
+  }
+
+ public:
+  // Used by frame functions to describe locations.
+  void describe(int owner, intptr_t* location, const char* description, int priority = 0);
+
+  bool validate();
+  void print();
+};
+
+#endif
 
 //
 // StackFrameStream iterates through the frames of a thread starting from
diff --git a/src/share/vm/runtime/thread.cpp b/src/share/vm/runtime/thread.cpp
index 6cb2deeb4..02f4e7f49 100644
--- a/src/share/vm/runtime/thread.cpp
+++ b/src/share/vm/runtime/thread.cpp
@@ -31,6 +31,7 @@
 #include "compiler/compileBroker.hpp"
 #include "interpreter/interpreter.hpp"
 #include "interpreter/linkResolver.hpp"
+#include "interpreter/oopMapCache.hpp"
 #include "jvmtifiles/jvmtiEnv.hpp"
 #include "memory/oopFactory.hpp"
 #include "memory/universe.inline.hpp"
@@ -2860,6 +2861,25 @@ void JavaThread::trace_frames() {
 }
 
 
+// Print or validate the layout of stack frames
+void JavaThread::print_frame_layout(int depth, bool validate_only) {
+  ResourceMark rm;
+  PRESERVE_EXCEPTION_MARK;
+  FrameValues values;
+  int frame_no = 0;
+  for(StackFrameStream fst(this, false); !fst.is_done(); fst.next()) {
+    fst.current()->describe(values, ++frame_no);
+    if (depth == frame_no) break;
+  }
+  if (validate_only) {
+    values.validate();
+  } else {
+    tty->print_cr("[Describe stack layout]");
+    values.print();
+  }
+}
+
+
 void JavaThread::trace_stack_from(vframe* start_vf) {
   ResourceMark rm;
   int vframe_no = 1;
diff --git a/src/share/vm/runtime/thread.hpp b/src/share/vm/runtime/thread.hpp
index 73e0fda02..eefe93bc7 100644
--- a/src/share/vm/runtime/thread.hpp
+++ b/src/share/vm/runtime/thread.hpp
@@ -1380,6 +1380,12 @@ public:
   void trace_stack_from(vframe* start_vf)        PRODUCT_RETURN;
   void trace_frames()                            PRODUCT_RETURN;
 
+  // Print an annotated view of the stack frames
+  void print_frame_layout(int depth = 0, bool validate_only = false) PRODUCT_RETURN;
+  void validate_frame_layout() {
+    print_frame_layout(0, true);
+  }
+
   // Returns the number of stack frames on the stack
   int depth() const;
 
diff --git a/src/share/vm/utilities/debug.cpp b/src/share/vm/utilities/debug.cpp
index cb5efc08d..27eaea8b4 100644
--- a/src/share/vm/utilities/debug.cpp
+++ b/src/share/vm/utilities/debug.cpp
@@ -469,6 +469,7 @@ extern "C" void verify() {
 extern "C" void pp(void* p) {
   Command c("pp");
   FlagSetting fl(PrintVMMessages, true);
+  FlagSetting f2(DisplayVMOutput, true);
   if (Universe::heap()->is_in(p)) {
     oop obj = oop(p);
     obj->print();
@@ -507,6 +508,17 @@ extern "C" void ps() { // print stack
 
 }
 
+extern "C" void pfl() {
+  // print frame layout
+  Command c("pfl");
+  JavaThread* p = JavaThread::active();
+  tty->print(" for thread: ");
+  p->print();
+  tty->cr();
+  if (p->has_last_Java_frame()) {
+    p->print_frame_layout();
+  }
+}
 
 extern "C" void psf() { // print stack frames
   {