8035647: PPC64: Support for elf v2 abi.

Summary: ELFv2 ABI used by the little endian PowerPC64 on Linux.
Reviewed-by: kvn
Contributed-by: asmundak@google.com

src/cpu/ppc/vm/assembler_ppc.hpp

@@ -124,6 +124,7 @@ class Argument VALUE_OBJ_CLASS_SPEC {
   }
 };
 
+#if !defined(ABI_ELFv2)
 // A ppc64 function descriptor.
 struct FunctionDescriptor VALUE_OBJ_CLASS_SPEC {
  private:
@@ -161,6 +162,7 @@ struct FunctionDescriptor VALUE_OBJ_CLASS_SPEC {
     _env   = (address) 0xbad;
   }
 };
+#endif
 
 class Assembler : public AbstractAssembler {
  protected:
@@ -1067,6 +1069,7 @@ class Assembler : public AbstractAssembler {
   // Emit an address.
   inline address emit_addr(const address addr = NULL);
 
+#if !defined(ABI_ELFv2)
   // Emit a function descriptor with the specified entry point, TOC,
   // and ENV. If the entry point is NULL, the descriptor will point
   // just past the descriptor.
@@ -1074,6 +1077,7 @@ class Assembler : public AbstractAssembler {
   inline address emit_fd(address entry = NULL,
                          address toc = (address) FunctionDescriptor::friend_toc,
                          address env = (address) FunctionDescriptor::friend_env);
+#endif
 
   /////////////////////////////////////////////////////////////////////////////////////
   // PPC instructions

src/cpu/ppc/vm/assembler_ppc.inline.hpp

@@ -55,6 +55,7 @@ inline address Assembler::emit_addr(const address addr) {
   return start;
 }
 
+#if !defined(ABI_ELFv2)
 // Emit a function descriptor with the specified entry point, TOC, and
 // ENV. If the entry point is NULL, the descriptor will point just
 // past the descriptor.
@@ -73,6 +74,7 @@ inline address Assembler::emit_fd(address entry, address toc, address env) {
 
   return (address)fd;
 }
+#endif
 
 // Issue an illegal instruction. 0 is guaranteed to be an illegal instruction.
 inline void Assembler::illtrap() { Assembler::emit_int32(0); }

src/cpu/ppc/vm/cppInterpreter_ppc.cpp

@@ -1136,7 +1136,9 @@ address CppInterpreterGenerator::generate_native_entry(void) {
   // (outgoing C args), R3_ARG1 to R10_ARG8, and F1_ARG1 to
   // F13_ARG13.
   __ mr(R3_ARG1, R18_locals);
+#if !defined(ABI_ELFv2)
   __ ld(signature_handler_fd, 0, signature_handler_fd);
+#endif
   __ call_stub(signature_handler_fd);
   // reload method
   __ ld(R19_method, state_(_method));
@@ -1295,8 +1297,13 @@ address CppInterpreterGenerator::generate_native_entry(void) {
   // native result acrosss the call. No oop is present
 
   __ mr(R3_ARG1, R16_thread);
+#if defined(ABI_ELFv2)
+  __ call_c(CAST_FROM_FN_PTR(address, JavaThread::check_special_condition_for_native_trans),
+            relocInfo::none);
+#else
   __ call_c(CAST_FROM_FN_PTR(FunctionDescriptor*, JavaThread::check_special_condition_for_native_trans),
             relocInfo::none);
+#endif
   __ bind(sync_check_done);
 
   //=============================================================================
@@ -1413,7 +1420,7 @@ address CppInterpreterGenerator::generate_native_entry(void) {
   // First, pop to caller's frame.
   __ pop_interpreter_frame(R11_scratch1, R12_scratch2, R21_tmp1  /* set to return pc */, R22_tmp2);
 
-  __ push_frame_abi112(0, R11_scratch1);
+  __ push_frame_reg_args(0, R11_scratch1);
   // Get the address of the exception handler.
   __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::exception_handler_for_return_address),
                   R16_thread,
@@ -2545,7 +2552,7 @@ address CppInterpreterGenerator::generate_normal_entry(void) {
   __ mr(R4_ARG2, R3_ARG1);  // ARG2 := ARG1
 
   // Find the address of the "catch_exception" stub.
-  __ push_frame_abi112(0, R11_scratch1);
+  __ push_frame_reg_args(0, R11_scratch1);
   __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::exception_handler_for_return_address),
                   R16_thread,
                   R4_ARG2);

src/cpu/ppc/vm/frame_ppc.hpp

@@ -50,7 +50,7 @@
   //            [C_FRAME]
   //
   //  C_FRAME:
-  //    0       [ABI_112]
+  //    0       [ABI_REG_ARGS]
   //    112     CARG_9: outgoing arg 9 (arg_1 ... arg_8 via gpr_3 ... gpr_{10})
   //            ...
   //    40+M*8  CARG_M: outgoing arg M (M is the maximum of outgoing args taken over all call sites in the procedure)
@@ -77,7 +77,7 @@
   //    32      reserved
   //    40      space for TOC (=R2) register for next call
   //
-  //  ABI_112:
+  //  ABI_REG_ARGS:
   //    0       [ABI_48]
   //    48      CARG_1: spill slot for outgoing arg 1. used by next callee.
   //    ...     ...
@@ -95,23 +95,25 @@
     log_2_of_alignment_in_bits = 7
   };
 
-  // ABI_48:
-  struct abi_48 {
+  // ABI_MINFRAME:
+  struct abi_minframe {
     uint64_t callers_sp;
     uint64_t cr;                                  //_16
     uint64_t lr;
+#if !defined(ABI_ELFv2)
     uint64_t reserved1;                           //_16
     uint64_t reserved2;
+#endif
     uint64_t toc;                                 //_16
     // nothing to add here!
     // aligned to frame::alignment_in_bytes (16)
   };
 
   enum {
-    abi_48_size = sizeof(abi_48)
+    abi_minframe_size = sizeof(abi_minframe)
   };
 
-  struct abi_112 : abi_48 {
+  struct abi_reg_args : abi_minframe {
     uint64_t carg_1;
     uint64_t carg_2;                              //_16
     uint64_t carg_3;
@@ -124,13 +126,13 @@
   };
 
   enum {
-    abi_112_size = sizeof(abi_112)
+    abi_reg_args_size = sizeof(abi_reg_args)
   };
 
   #define _abi(_component) \
-          (offset_of(frame::abi_112, _component))
+          (offset_of(frame::abi_reg_args, _component))
 
-  struct abi_112_spill : abi_112 {
+  struct abi_reg_args_spill : abi_reg_args {
     // additional spill slots
     uint64_t spill_ret;
     uint64_t spill_fret;                          //_16
@@ -138,11 +140,11 @@
   };
 
   enum {
-    abi_112_spill_size = sizeof(abi_112_spill)
+    abi_reg_args_spill_size = sizeof(abi_reg_args_spill)
   };
 
-  #define _abi_112_spill(_component) \
-          (offset_of(frame::abi_112_spill, _component))
+  #define _abi_reg_args_spill(_component) \
+          (offset_of(frame::abi_reg_args_spill, _component))
 
   // non-volatile GPRs:
 
@@ -242,7 +244,7 @@
   //            [ENTRY_FRAME_LOCALS]
   //
   //  PARENT_IJAVA_FRAME_ABI:
-  //    0       [ABI_48]
+  //    0       [ABI_MINFRAME]
   //            top_frame_sp
   //            initial_caller_sp
   //
@@ -258,7 +260,7 @@
 
   // PARENT_IJAVA_FRAME_ABI
 
-  struct parent_ijava_frame_abi : abi_48 {
+  struct parent_ijava_frame_abi : abi_minframe {
     // SOE registers.
     // C2i adapters spill their top-frame stack-pointer here.
     uint64_t top_frame_sp;                        //      carg_1
@@ -285,7 +287,7 @@
     uint64_t carg_6_unused;                       //_16   carg_6
     uint64_t carg_7_unused;                       //      carg_7
     // Use arg8 for storing frame_manager_lr. The size of
-    // top_ijava_frame_abi must match abi_112.
+    // top_ijava_frame_abi must match abi_reg_args.
     uint64_t frame_manager_lr;                    //_16   carg_8
     // nothing to add here!
     // aligned to frame::alignment_in_bytes (16)
@@ -395,8 +397,8 @@
   intptr_t* fp() const { return _fp; }
 
   // Accessors for ABIs
-  inline abi_48* own_abi()     const { return (abi_48*) _sp; }
-  inline abi_48* callers_abi() const { return (abi_48*) _fp; }
+  inline abi_minframe* own_abi()     const { return (abi_minframe*) _sp; }
+  inline abi_minframe* callers_abi() const { return (abi_minframe*) _fp; }
 
  private:
 

src/cpu/ppc/vm/interpreterRT_ppc.cpp

@@ -109,8 +109,10 @@ void InterpreterRuntime::SignatureHandlerGenerator::pass_object() {
 }
 
 void InterpreterRuntime::SignatureHandlerGenerator::generate(uint64_t fingerprint) {
+#if !defined(ABI_ELFv2)
   // Emit fd for current codebuffer. Needs patching!
   __ emit_fd();
+#endif
 
   // Generate code to handle arguments.
   iterate(fingerprint);
@@ -127,11 +129,13 @@ void InterpreterRuntime::SignatureHandlerGenerator::generate(uint64_t fingerprin
 // Implementation of SignatureHandlerLibrary
 
 void SignatureHandlerLibrary::pd_set_handler(address handler) {
+#if !defined(ABI_ELFv2)
   // patch fd here.
   FunctionDescriptor* fd = (FunctionDescriptor*) handler;
 
   fd->set_entry(handler + (int)sizeof(FunctionDescriptor));
   assert(fd->toc() == (address)0xcafe, "need to adjust TOC here");
+#endif
 }
 
 

src/cpu/ppc/vm/interpreter_ppc.cpp

@@ -128,13 +128,13 @@ address AbstractInterpreterGenerator::generate_slow_signature_handler() {
   const Register target_sp      = R28_tmp8;
   const FloatRegister floatSlot = F0;
 
-  address entry = __ emit_fd();
+  address entry = __ function_entry();
 
   __ save_LR_CR(R0);
   __ save_nonvolatile_gprs(R1_SP, _spill_nonvolatiles_neg(r14));
   // We use target_sp for storing arguments in the C frame.
   __ mr(target_sp, R1_SP);
-  __ push_frame_abi112_nonvolatiles(0, R11_scratch1);
+  __ push_frame_reg_args_nonvolatiles(0, R11_scratch1);
 
   __ mr(arg_java, R3_ARG1);
 
@@ -474,7 +474,7 @@ address InterpreterGenerator::generate_abstract_entry(void) {
 
   // Push a new C frame and save LR.
   __ save_LR_CR(R0);
-  __ push_frame_abi112(0, R11_scratch1);
+  __ push_frame_reg_args(0, R11_scratch1);
 
   // This is not a leaf but we have a JavaFrameAnchor now and we will
   // check (create) exceptions afterward so this is ok.

src/cpu/ppc/vm/macroAssembler_ppc.cpp

@@ -594,7 +594,13 @@ void MacroAssembler::bxx64_patchable(address dest, relocInfo::relocType rt, bool
            "can't identify emitted call");
   } else {
     // variant 1:
-
+#if defined(ABI_ELFv2)
+    nop();
+    calculate_address_from_global_toc(R12, dest, true, true, false);
+    mtctr(R12);
+    nop();
+    nop();
+#else
     mr(R0, R11);  // spill R11 -> R0.
 
     // Load the destination address into CTR,
@@ -604,6 +610,7 @@ void MacroAssembler::bxx64_patchable(address dest, relocInfo::relocType rt, bool
     mtctr(R11);
     mr(R11, R0);  // spill R11 <- R0.
     nop();
+#endif
 
     // do the call/jump
     if (link) {
@@ -912,16 +919,16 @@ void MacroAssembler::push_frame(unsigned int bytes, Register tmp) {
   }
 }
 
-// Push a frame of size `bytes' plus abi112 on top.
-void MacroAssembler::push_frame_abi112(unsigned int bytes, Register tmp) {
-  push_frame(bytes + frame::abi_112_size, tmp);
+// Push a frame of size `bytes' plus abi_reg_args on top.
+void MacroAssembler::push_frame_reg_args(unsigned int bytes, Register tmp) {
+  push_frame(bytes + frame::abi_reg_args_size, tmp);
 }
 
 // Setup up a new C frame with a spill area for non-volatile GPRs and
 // additional space for local variables.
-void MacroAssembler::push_frame_abi112_nonvolatiles(unsigned int bytes,
-                                                    Register tmp) {
-  push_frame(bytes + frame::abi_112_size + frame::spill_nonvolatiles_size, tmp);
+void MacroAssembler::push_frame_reg_args_nonvolatiles(unsigned int bytes,
+                                                      Register tmp) {
+  push_frame(bytes + frame::abi_reg_args_size + frame::spill_nonvolatiles_size, tmp);
 }
 
 // Pop current C frame.
@@ -929,6 +936,42 @@ void MacroAssembler::pop_frame() {
   ld(R1_SP, _abi(callers_sp), R1_SP);
 }
 
+#if defined(ABI_ELFv2)
+address MacroAssembler::branch_to(Register r_function_entry, bool and_link) {
+  // TODO(asmundak): make sure the caller uses R12 as function descriptor
+  // most of the times.
+  if (R12 != r_function_entry) {
+    mr(R12, r_function_entry);
+  }
+  mtctr(R12);
+  // Do a call or a branch.
+  if (and_link) {
+    bctrl();
+  } else {
+    bctr();
+  }
+  _last_calls_return_pc = pc();
+
+  return _last_calls_return_pc;
+}
+
+// Call a C function via a function descriptor and use full C
+// calling conventions. Updates and returns _last_calls_return_pc.
+address MacroAssembler::call_c(Register r_function_entry) {
+  return branch_to(r_function_entry, /*and_link=*/true);
+}
+
+// For tail calls: only branch, don't link, so callee returns to caller of this function.
+address MacroAssembler::call_c_and_return_to_caller(Register r_function_entry) {
+  return branch_to(r_function_entry, /*and_link=*/false);
+}
+
+address MacroAssembler::call_c(address function_entry, relocInfo::relocType rt) {
+  load_const(R12, function_entry, R0);
+  return branch_to(R12,  /*and_link=*/true);
+}
+
+#else
 // Generic version of a call to C function via a function descriptor
 // with variable support for C calling conventions (TOC, ENV, etc.).
 // Updates and returns _last_calls_return_pc.
@@ -1077,6 +1120,7 @@ address MacroAssembler::call_c_using_toc(const FunctionDescriptor* fd,
   }
   return _last_calls_return_pc;
 }
+#endif
 
 void MacroAssembler::call_VM_base(Register oop_result,
                                   Register last_java_sp,
@@ -1091,8 +1135,11 @@ void MacroAssembler::call_VM_base(Register oop_result,
 
   // ARG1 must hold thread address.
   mr(R3_ARG1, R16_thread);
-
+#if defined(ABI_ELFv2)
+  address return_pc = call_c(entry_point, relocInfo::none);
+#else
   address return_pc = call_c((FunctionDescriptor*)entry_point, relocInfo::none);
+#endif
 
   reset_last_Java_frame();
 
@@ -1113,7 +1160,11 @@ void MacroAssembler::call_VM_base(Register oop_result,
 
 void MacroAssembler::call_VM_leaf_base(address entry_point) {
   BLOCK_COMMENT("call_VM_leaf {");
+#if defined(ABI_ELFv2)
+  call_c(entry_point, relocInfo::none);
+#else
   call_c(CAST_FROM_FN_PTR(FunctionDescriptor*, entry_point), relocInfo::none);
+#endif
   BLOCK_COMMENT("} call_VM_leaf");
 }
 
@@ -2227,7 +2278,7 @@ void MacroAssembler::g1_write_barrier_pre(Register Robj, RegisterOrConstant offs
   // VM call need frame to access(write) O register.
   if (needs_frame) {
     save_LR_CR(Rtmp1);
-    push_frame_abi112(0, Rtmp2);
+    push_frame_reg_args(0, Rtmp2);
   }
 
   if (Rpre_val->is_volatile() && Robj == noreg) mr(R31, Rpre_val); // Save pre_val across C call if it was preloaded.
@@ -3006,13 +3057,13 @@ void MacroAssembler::verify_oop(Register oop, const char* msg) {
   mr(R0, tmp);
   // kill tmp
   save_LR_CR(tmp);
-  push_frame_abi112(nbytes_save, tmp);
+  push_frame_reg_args(nbytes_save, tmp);
   // restore tmp
   mr(tmp, R0);
   save_volatile_gprs(R1_SP, 112); // except R0
-  // load FunctionDescriptor**
+  // load FunctionDescriptor** / entry_address *
   load_const(tmp, fd);
-  // load FunctionDescriptor*
+  // load FunctionDescriptor* / entry_address
   ld(tmp, 0, tmp);
   mr(R4_ARG2, oop);
   load_const(R3_ARG1, (address)msg);

src/cpu/ppc/vm/macroAssembler_ppc.hpp

@@ -279,12 +279,12 @@ class MacroAssembler: public Assembler {
   // Push a frame of size `bytes'. No abi space provided.
   void push_frame(unsigned int bytes, Register tmp);
 
-  // Push a frame of size `bytes' plus abi112 on top.
-  void push_frame_abi112(unsigned int bytes, Register tmp);
+  // Push a frame of size `bytes' plus abi_reg_args on top.
+  void push_frame_reg_args(unsigned int bytes, Register tmp);
 
   // Setup up a new C frame with a spill area for non-volatile GPRs and additional
   // space for local variables
-  void push_frame_abi112_nonvolatiles(unsigned int bytes, Register tmp);
+  void push_frame_reg_args_nonvolatiles(unsigned int bytes, Register tmp);
 
   // pop current C frame
   void pop_frame();
@@ -296,17 +296,31 @@ class MacroAssembler: public Assembler {
  private:
   address _last_calls_return_pc;
 
+#if defined(ABI_ELFv2)
+  // Generic version of a call to C function.
+  // Updates and returns _last_calls_return_pc.
+  address branch_to(Register function_entry, bool and_link);
+#else
   // Generic version of a call to C function via a function descriptor
   // with variable support for C calling conventions (TOC, ENV, etc.).
   // updates and returns _last_calls_return_pc.
   address branch_to(Register function_descriptor, bool and_link, bool save_toc_before_call,
                     bool restore_toc_after_call, bool load_toc_of_callee, bool load_env_of_callee);
+#endif
 
  public:
 
   // Get the pc where the last call will return to. returns _last_calls_return_pc.
   inline address last_calls_return_pc();
 
+#if defined(ABI_ELFv2)
+  // Call a C function via a function descriptor and use full C
+  // calling conventions. Updates and returns _last_calls_return_pc.
+  address call_c(Register function_entry);
+  // For tail calls: only branch, don't link, so callee returns to caller of this function.
+  address call_c_and_return_to_caller(Register function_entry);
+  address call_c(address function_entry, relocInfo::relocType rt);
+#else
   // Call a C function via a function descriptor and use full C
   // calling conventions. Updates and returns _last_calls_return_pc.
   address call_c(Register function_descriptor);
@@ -315,6 +329,7 @@ class MacroAssembler: public Assembler {
   address call_c(const FunctionDescriptor* function_descriptor, relocInfo::relocType rt);
   address call_c_using_toc(const FunctionDescriptor* function_descriptor, relocInfo::relocType rt,
                            Register toc);
+#endif
 
  protected:
 
@@ -649,6 +664,11 @@ class MacroAssembler: public Assembler {
   void _verify_method_ptr(Register reg, const char * msg, const char * file, int line) {}
   void _verify_klass_ptr(Register reg, const char * msg, const char * file, int line) {}
 
+  // Convenience method returning function entry. For the ELFv1 case
+  // creates function descriptor at the current address and returs
+  // the pointer to it. For the ELFv2 case returns the current address.
+  inline address function_entry();
+
 #define verify_method_ptr(reg) _verify_method_ptr(reg, "broken method " #reg, __FILE__, __LINE__)
 #define verify_klass_ptr(reg) _verify_klass_ptr(reg, "broken klass " #reg, __FILE__, __LINE__)
 

src/cpu/ppc/vm/macroAssembler_ppc.inline.hpp

@@ -385,4 +385,10 @@ inline void MacroAssembler::trap_range_check_ge(Register a, int si16) {
   twi(traptoEqual | traptoGreaterThanUnsigned, a/*reg a*/, si16);
 }
 
+#if defined(ABI_ELFv2)
+inline address MacroAssembler::function_entry() { return pc(); }
+#else
+inline address MacroAssembler::function_entry() { return emit_fd(); }
+#endif
+
 #endif // CPU_PPC_VM_MACROASSEMBLER_PPC_INLINE_HPP

src/cpu/ppc/vm/methodHandles_ppc.cpp

@@ -453,11 +453,11 @@ void trace_method_handle_stub(const char* adaptername,
 
   if (Verbose) {
     tty->print_cr("Registers:");
-    const int abi_offset = frame::abi_112_size / 8;
+    const int abi_offset = frame::abi_reg_args_size / 8;
     for (int i = R3->encoding(); i <= R12->encoding(); i++) {
       Register r = as_Register(i);
       int count = i - R3->encoding();
-      // The registers are stored in reverse order on the stack (by save_volatile_gprs(R1_SP, abi_112_size)).
+      // The registers are stored in reverse order on the stack (by save_volatile_gprs(R1_SP, abi_reg_args_size)).
       tty->print("%3s=" PTR_FORMAT, r->name(), saved_regs[abi_offset + count]);
       if ((count + 1) % 4 == 0) {
         tty->cr();
@@ -524,9 +524,9 @@ void MethodHandles::trace_method_handle(MacroAssembler* _masm, const char* adapt
   __ save_LR_CR(R0);
   __ mr(R0, R1_SP);                     // saved_sp
   assert(Assembler::is_simm(-nbytes_save, 16), "Overwriting R0");
-  // push_frame_abi112 only uses R0 if nbytes_save is wider than 16 bit
-  __ push_frame_abi112(nbytes_save, R0);
-  __ save_volatile_gprs(R1_SP, frame::abi_112_size); // Except R0.
+  // Push_frame_reg_args only uses R0 if nbytes_save is wider than 16 bit.
+  __ push_frame_reg_args(nbytes_save, R0);
+  __ save_volatile_gprs(R1_SP, frame::abi_reg_args_size); // Except R0.
 
   __ load_const(R3_ARG1, (address)adaptername);
   __ mr(R4_ARG2, R23_method_handle);

src/cpu/ppc/vm/ppc.ad

@@ -1008,7 +1008,11 @@ int MachCallDynamicJavaNode::ret_addr_offset() {
 }
 
 int MachCallRuntimeNode::ret_addr_offset() {
+#if defined(ABI_ELFv2)
+  return 28;
+#else
   return 40;
+#endif
 }
 
 //=============================================================================
@@ -3686,6 +3690,10 @@ encode %{
     MacroAssembler _masm(&cbuf);
     const address start_pc = __ pc();
 
+#if defined(ABI_ELFv2)
+    address entry= !($meth$$method) ? NULL : (address)$meth$$method;
+    __ call_c(entry, relocInfo::runtime_call_type);
+#else
     // The function we're going to call.
     FunctionDescriptor fdtemp;
     const FunctionDescriptor* fd = !($meth$$method) ? &fdtemp : (FunctionDescriptor*)$meth$$method;
@@ -3696,6 +3704,7 @@ encode %{
     // Put entry, env, toc into the constant pool, this needs up to 3 constant
     // pool entries; call_c_using_toc will optimize the call.
     __ call_c_using_toc(fd, relocInfo::runtime_call_type, Rtoc);
+#endif
 
     // Check the ret_addr_offset.
     assert(((MachCallRuntimeNode*)this)->ret_addr_offset() ==  __ last_calls_return_pc() - start_pc,
@@ -3711,20 +3720,25 @@ encode %{
     __ mtctr($src$$Register);
   %}
 
-  // postalloc expand emitter for runtime leaf calls.
+  // Postalloc expand emitter for runtime leaf calls.
   enc_class postalloc_expand_java_to_runtime_call(method meth, iRegLdst toc) %{
+    loadConLNodesTuple loadConLNodes_Entry;
+#if defined(ABI_ELFv2)
+    jlong entry_address = (jlong) this->entry_point();
+    assert(entry_address, "need address here");
+    loadConLNodes_Entry = loadConLNodesTuple_create(C, ra_, n_toc, new (C) immLOper(entry_address),
+                                                    OptoReg::Name(R12_H_num), OptoReg::Name(R12_num));
+#else
     // Get the struct that describes the function we are about to call.
     FunctionDescriptor* fd = (FunctionDescriptor*) this->entry_point();
     assert(fd, "need fd here");
+    jlong entry_address = (jlong) fd->entry();
     // new nodes
-    loadConLNodesTuple loadConLNodes_Entry;
     loadConLNodesTuple loadConLNodes_Env;
     loadConLNodesTuple loadConLNodes_Toc;
-    MachNode         *mtctr = NULL;
-    MachCallLeafNode *call  = NULL;
 
     // Create nodes and operands for loading the entry point.
-    loadConLNodes_Entry = loadConLNodesTuple_create(C, ra_, n_toc, new (C) immLOper((jlong) fd->entry()),
+    loadConLNodes_Entry = loadConLNodesTuple_create(C, ra_, n_toc, new (C) immLOper(entry_address),
                                                     OptoReg::Name(R12_H_num), OptoReg::Name(R12_num));
 
 
@@ -3745,8 +3759,9 @@ encode %{
     // Create nodes and operands for loading the Toc point.
     loadConLNodes_Toc = loadConLNodesTuple_create(C, ra_, n_toc, new (C) immLOper((jlong) fd->toc()),
                                                   OptoReg::Name(R2_H_num), OptoReg::Name(R2_num));
+#endif // ABI_ELFv2
     // mtctr node
-    mtctr = new (C) CallLeafDirect_mtctrNode();
+    MachNode *mtctr = new (C) CallLeafDirect_mtctrNode();
 
     assert(loadConLNodes_Entry._last != NULL, "entry must exist");
     mtctr->add_req(0, loadConLNodes_Entry._last);
@@ -3755,10 +3770,10 @@ encode %{
     mtctr->_opnds[1] = new (C) iRegLdstOper();
 
     // call node
-    call = new (C) CallLeafDirectNode();
+    MachCallLeafNode *call = new (C) CallLeafDirectNode();
 
     call->_opnds[0] = _opnds[0];
-    call->_opnds[1] = new (C) methodOper((intptr_t) fd->entry()); // may get set later
+    call->_opnds[1] = new (C) methodOper((intptr_t) entry_address); // May get set later.
 
     // Make the new call node look like the old one.
     call->_name        = _name;
@@ -3785,8 +3800,10 @@ encode %{
     // These must be reqired edges, as the registers are live up to
     // the call. Else the constants are handled as kills.
     call->add_req(mtctr);
+#if !defined(ABI_ELFv2)
     call->add_req(loadConLNodes_Env._last);
     call->add_req(loadConLNodes_Toc._last);
+#endif
 
     // ...as well as prec
     for (uint i = req(); i < len(); ++i) {
@@ -3799,10 +3816,12 @@ encode %{
     // Insert the new nodes.
     if (loadConLNodes_Entry._large_hi) nodes->push(loadConLNodes_Entry._large_hi);
     if (loadConLNodes_Entry._last)     nodes->push(loadConLNodes_Entry._last);
+#if !defined(ABI_ELFv2)
     if (loadConLNodes_Env._large_hi)   nodes->push(loadConLNodes_Env._large_hi);
     if (loadConLNodes_Env._last)       nodes->push(loadConLNodes_Env._last);
     if (loadConLNodes_Toc._large_hi)   nodes->push(loadConLNodes_Toc._large_hi);
     if (loadConLNodes_Toc._last)       nodes->push(loadConLNodes_Toc._last);
+#endif
     nodes->push(mtctr);
     nodes->push(call);
   %}
@@ -3849,7 +3868,7 @@ frame %{
   // out_preserve_stack_slots for calls to C. Supports the var-args
   // backing area for register parms.
   //
-  varargs_C_out_slots_killed(((frame::abi_112_size - frame::jit_out_preserve_size) / VMRegImpl::stack_slot_size));
+  varargs_C_out_slots_killed(((frame::abi_reg_args_size - frame::jit_out_preserve_size) / VMRegImpl::stack_slot_size));
 
   // The after-PROLOG location of the return address. Location of
   // return address specifies a type (REG or STACK) and a number

src/cpu/ppc/vm/runtime_ppc.cpp

@@ -87,7 +87,7 @@ void OptoRuntime::generate_exception_blob() {
 
   address start = __ pc();
 
-  int frame_size_in_bytes = frame::abi_112_size;
+  int frame_size_in_bytes = frame::abi_reg_args_size;
   OopMap* map = new OopMap(frame_size_in_bytes / sizeof(jint), 0);
 
   // Exception pc is 'return address' for stack walker.
@@ -99,7 +99,7 @@ void OptoRuntime::generate_exception_blob() {
 
   // Save callee-saved registers.
   // Push a C frame for the exception blob. It is needed for the C call later on.
-  __ push_frame_abi112(0, R11_scratch1);
+  __ push_frame_reg_args(0, R11_scratch1);
 
   // This call does all the hard work. It checks if an exception handler
   // exists in the method.
@@ -109,8 +109,12 @@ void OptoRuntime::generate_exception_blob() {
   __ set_last_Java_frame(/*sp=*/R1_SP, noreg);
 
   __ mr(R3_ARG1, R16_thread);
+#if defined(ABI_ELFv2)
+  __ call_c((address) OptoRuntime::handle_exception_C, relocInfo::none);
+#else
   __ call_c(CAST_FROM_FN_PTR(FunctionDescriptor*, OptoRuntime::handle_exception_C),
             relocInfo::none);
+#endif
   address calls_return_pc = __ last_calls_return_pc();
 # ifdef ASSERT
   __ cmpdi(CCR0, R3_RET, 0);
@@ -162,7 +166,11 @@ void OptoRuntime::generate_exception_blob() {
   __ bind(mh_callsite);
   __ mr(R31, R3_RET); // Save branch address.
   __ mr(R3_ARG1, R16_thread);
+#if defined(ABI_ELFv2)
+  __ call_c((address) adjust_SP_for_methodhandle_callsite, relocInfo::none);
+#else
   __ call_c(CAST_FROM_FN_PTR(FunctionDescriptor*, adjust_SP_for_methodhandle_callsite), relocInfo::none);
+#endif
   // Returns unextended_sp in R3_RET.
 
   __ mtctr(R31); // Move address of exception handler to SR_CTR.

src/cpu/ppc/vm/sharedRuntime_ppc.cpp

@@ -67,7 +67,7 @@ class RegisterSaver {
     return_pc_is_thread_saved_exception_pc
   };
 
-  static OopMap* push_frame_abi112_and_save_live_registers(MacroAssembler* masm,
+  static OopMap* push_frame_reg_args_and_save_live_registers(MacroAssembler* masm,
                          int* out_frame_size_in_bytes,
                          bool generate_oop_map,
                          int return_pc_adjustment,
@@ -200,12 +200,12 @@ static const RegisterSaver::LiveRegType RegisterSaver_LiveRegs[] = {
   RegisterSaver_LiveIntReg(   R30 ), // r30 must be the last register
 };
 
-OopMap* RegisterSaver::push_frame_abi112_and_save_live_registers(MacroAssembler* masm,
+OopMap* RegisterSaver::push_frame_reg_args_and_save_live_registers(MacroAssembler* masm,
                          int* out_frame_size_in_bytes,
                          bool generate_oop_map,
                          int return_pc_adjustment,
                          ReturnPCLocation return_pc_location) {
-  // Push an abi112-frame and store all registers which may be live.
+  // Push an abi_reg_args-frame and store all registers which may be live.
   // If requested, create an OopMap: Record volatile registers as
   // callee-save values in an OopMap so their save locations will be
   // propagated to the RegisterMap of the caller frame during
@@ -221,7 +221,7 @@ OopMap* RegisterSaver::push_frame_abi112_and_save_live_registers(MacroAssembler*
                                    sizeof(RegisterSaver::LiveRegType);
   const int register_save_size   = regstosave_num * reg_size;
   const int frame_size_in_bytes  = round_to(register_save_size, frame::alignment_in_bytes)
-                                   + frame::abi_112_size;
+                                   + frame::abi_reg_args_size;
   *out_frame_size_in_bytes       = frame_size_in_bytes;
   const int frame_size_in_slots  = frame_size_in_bytes / sizeof(jint);
   const int register_save_offset = frame_size_in_bytes - register_save_size;
@@ -229,7 +229,7 @@ OopMap* RegisterSaver::push_frame_abi112_and_save_live_registers(MacroAssembler*
   // OopMap frame size is in c2 stack slots (sizeof(jint)) not bytes or words.
   OopMap* map = generate_oop_map ? new OopMap(frame_size_in_slots, 0) : NULL;
 
-  BLOCK_COMMENT("push_frame_abi112_and_save_live_registers {");
+  BLOCK_COMMENT("push_frame_reg_args_and_save_live_registers {");
 
   // Save r30 in the last slot of the not yet pushed frame so that we
   // can use it as scratch reg.
@@ -294,7 +294,7 @@ OopMap* RegisterSaver::push_frame_abi112_and_save_live_registers(MacroAssembler*
     offset += reg_size;
   }
 
-  BLOCK_COMMENT("} push_frame_abi112_and_save_live_registers");
+  BLOCK_COMMENT("} push_frame_reg_args_and_save_live_registers");
 
   // And we're done.
   return map;
@@ -699,15 +699,19 @@ int SharedRuntime::c_calling_convention(const BasicType *sig_bt,
 
   int i;
   VMReg reg;
-  // Leave room for C-compatible ABI_112.
-  int stk = (frame::abi_112_size - frame::jit_out_preserve_size) / VMRegImpl::stack_slot_size;
+  // Leave room for C-compatible ABI_REG_ARGS.
+  int stk = (frame::abi_reg_args_size - frame::jit_out_preserve_size) / VMRegImpl::stack_slot_size;
   int arg = 0;
   int freg = 0;
 
   // Avoid passing C arguments in the wrong stack slots.
+#if defined(ABI_ELFv2)
+  assert((SharedRuntime::out_preserve_stack_slots() + stk) * VMRegImpl::stack_slot_size == 96,
+         "passing C arguments in wrong stack slots");
+#else
   assert((SharedRuntime::out_preserve_stack_slots() + stk) * VMRegImpl::stack_slot_size == 112,
          "passing C arguments in wrong stack slots");
-
+#endif
   // We fill-out regs AND regs2 if an argument must be passed in a
   // register AND in a stack slot. If regs2 is NULL in such a
   // situation, we bail-out with a fatal error.
@@ -1504,7 +1508,11 @@ static void check_needs_gc_for_critical_native(MacroAssembler* masm,
 
   __ block_comment("block_for_jni_critical");
   address entry_point = CAST_FROM_FN_PTR(address, SharedRuntime::block_for_jni_critical);
+#if defined(ABI_ELFv2)
+  __ call_c(entry_point, relocInfo::runtime_call_type);
+#else
   __ call_c(CAST_FROM_FN_PTR(FunctionDescriptor*, entry_point), relocInfo::runtime_call_type);
+#endif
   address start           = __ pc() - __ offset(),
           calls_return_pc = __ last_calls_return_pc();
   oop_maps->add_gc_map(calls_return_pc - start, map);
@@ -1877,7 +1885,7 @@ nmethod *SharedRuntime::generate_native_wrapper(MacroAssembler *masm,
   // Layout of the native wrapper frame:
   // (stack grows upwards, memory grows downwards)
   //
-  // NW     [ABI_112]                  <-- 1) R1_SP
+  // NW     [ABI_REG_ARGS]             <-- 1) R1_SP
   //        [outgoing arguments]       <-- 2) R1_SP + out_arg_slot_offset
   //        [oopHandle area]           <-- 3) R1_SP + oop_handle_offset (save area for critical natives)
   //        klass                      <-- 4) R1_SP + klass_offset
@@ -2211,8 +2219,8 @@ nmethod *SharedRuntime::generate_native_wrapper(MacroAssembler *masm,
     // slow case of monitor enter. Inline a special case of call_VM that
     // disallows any pending_exception.
 
-    // Save argument registers and leave room for C-compatible ABI_112.
-    int frame_size = frame::abi_112_size +
+    // Save argument registers and leave room for C-compatible ABI_REG_ARGS.
+    int frame_size = frame::abi_reg_args_size +
                      round_to(total_c_args * wordSize, frame::alignment_in_bytes);
     __ mr(R11_scratch1, R1_SP);
     RegisterSaver::push_frame_and_save_argument_registers(masm, R12_scratch2, frame_size, total_c_args, out_regs, out_regs2);
@@ -2250,9 +2258,12 @@ nmethod *SharedRuntime::generate_native_wrapper(MacroAssembler *masm,
 
   // The JNI call
   // --------------------------------------------------------------------------
-
+#if defined(ABI_ELFv2)
+  __ call_c(native_func, relocInfo::runtime_call_type);
+#else
   FunctionDescriptor* fd_native_method = (FunctionDescriptor*) native_func;
   __ call_c(fd_native_method, relocInfo::runtime_call_type);
+#endif
 
 
   // Now, we are back from the native code.
@@ -2724,7 +2735,7 @@ void SharedRuntime::generate_deopt_blob() {
   OopMapSet *oop_maps = new OopMapSet();
 
   // size of ABI112 plus spill slots for R3_RET and F1_RET.
-  const int frame_size_in_bytes = frame::abi_112_spill_size;
+  const int frame_size_in_bytes = frame::abi_reg_args_spill_size;
   const int frame_size_in_slots = frame_size_in_bytes / sizeof(jint);
   int first_frame_size_in_bytes = 0; // frame size of "unpack frame" for call to fetch_unroll_info.
 
@@ -2757,11 +2768,11 @@ void SharedRuntime::generate_deopt_blob() {
 
   // Push the "unpack frame"
   // Save everything in sight.
-  map = RegisterSaver::push_frame_abi112_and_save_live_registers(masm,
-                                                                 &first_frame_size_in_bytes,
-                                                                 /*generate_oop_map=*/ true,
-                                                                 return_pc_adjustment_no_exception,
-                                                                 RegisterSaver::return_pc_is_lr);
+  map = RegisterSaver::push_frame_reg_args_and_save_live_registers(masm,
+                                                                   &first_frame_size_in_bytes,
+                                                                   /*generate_oop_map=*/ true,
+                                                                   return_pc_adjustment_no_exception,
+                                                                   RegisterSaver::return_pc_is_lr);
   assert(map != NULL, "OopMap must have been created");
 
   __ li(exec_mode_reg, Deoptimization::Unpack_deopt);
@@ -2787,11 +2798,11 @@ void SharedRuntime::generate_deopt_blob() {
   // Push the "unpack frame".
   // Save everything in sight.
   assert(R4 == R4_ARG2, "exception pc must be in r4");
-  RegisterSaver::push_frame_abi112_and_save_live_registers(masm,
-                                                           &first_frame_size_in_bytes,
-                                                           /*generate_oop_map=*/ false,
-                                                           return_pc_adjustment_exception,
-                                                           RegisterSaver::return_pc_is_r4);
+  RegisterSaver::push_frame_reg_args_and_save_live_registers(masm,
+                                                             &first_frame_size_in_bytes,
+                                                             /*generate_oop_map=*/ false,
+                                                             return_pc_adjustment_exception,
+                                                             RegisterSaver::return_pc_is_r4);
 
   // Deopt during an exception. Save exec mode for unpack_frames.
   __ li(exec_mode_reg, Deoptimization::Unpack_exception);
@@ -2876,8 +2887,8 @@ void SharedRuntime::generate_deopt_blob() {
   // ...).
 
   // Spill live volatile registers since we'll do a call.
-  __ std( R3_RET,  _abi_112_spill(spill_ret),  R1_SP);
-  __ stfd(F1_RET, _abi_112_spill(spill_fret), R1_SP);
+  __ std( R3_RET, _abi_reg_args_spill(spill_ret),  R1_SP);
+  __ stfd(F1_RET, _abi_reg_args_spill(spill_fret), R1_SP);
 
   // Let the unpacker layout information in the skeletal frames just
   // allocated.
@@ -2889,8 +2900,8 @@ void SharedRuntime::generate_deopt_blob() {
   __ reset_last_Java_frame();
 
   // Restore the volatiles saved above.
-  __ ld( R3_RET, _abi_112_spill(spill_ret),  R1_SP);
-  __ lfd(F1_RET, _abi_112_spill(spill_fret), R1_SP);
+  __ ld( R3_RET, _abi_reg_args_spill(spill_ret),  R1_SP);
+  __ lfd(F1_RET, _abi_reg_args_spill(spill_fret), R1_SP);
 
   // Pop the unpack frame.
   __ pop_frame();
@@ -2930,7 +2941,7 @@ void SharedRuntime::generate_uncommon_trap_blob() {
   Register unc_trap_reg     = R23_tmp3;
 
   OopMapSet* oop_maps = new OopMapSet();
-  int frame_size_in_bytes = frame::abi_112_size;
+  int frame_size_in_bytes = frame::abi_reg_args_size;
   OopMap* map = new OopMap(frame_size_in_bytes / sizeof(jint), 0);
 
   // stack: (deoptee, optional i2c, caller_of_deoptee, ...).
@@ -2943,7 +2954,7 @@ void SharedRuntime::generate_uncommon_trap_blob() {
   __ save_LR_CR(R11_scratch1);
 
   // Push an "uncommon_trap" frame.
-  __ push_frame_abi112(0, R11_scratch1);
+  __ push_frame_reg_args(0, R11_scratch1);
 
   // stack: (unpack frame, deoptee, optional i2c, caller_of_deoptee, ...).
 
@@ -2996,7 +3007,7 @@ void SharedRuntime::generate_uncommon_trap_blob() {
   // interpreter frames just created.
 
   // Push a simple "unpack frame" here.
-  __ push_frame_abi112(0, R11_scratch1);
+  __ push_frame_reg_args(0, R11_scratch1);
 
   // stack: (unpack frame, skeletal interpreter frame, ..., optional
   // skeletal interpreter frame, optional c2i, caller of deoptee,
@@ -3064,11 +3075,11 @@ SafepointBlob* SharedRuntime::generate_handler_blob(address call_ptr, int poll_t
   }
 
   // Save registers, fpu state, and flags.
-  map = RegisterSaver::push_frame_abi112_and_save_live_registers(masm,
-                                                                 &frame_size_in_bytes,
-                                                                 /*generate_oop_map=*/ true,
-                                                                 /*return_pc_adjustment=*/0,
-                                                                 return_pc_location);
+  map = RegisterSaver::push_frame_reg_args_and_save_live_registers(masm,
+                                                                   &frame_size_in_bytes,
+                                                                   /*generate_oop_map=*/ true,
+                                                                   /*return_pc_adjustment=*/0,
+                                                                   return_pc_location);
 
   // The following is basically a call_VM. However, we need the precise
   // address of the call in order to generate an oopmap. Hence, we do all the
@@ -3151,11 +3162,11 @@ RuntimeStub* SharedRuntime::generate_resolve_blob(address destination, const cha
 
   address start = __ pc();
 
-  map = RegisterSaver::push_frame_abi112_and_save_live_registers(masm,
-                                                                 &frame_size_in_bytes,
-                                                                 /*generate_oop_map*/ true,
-                                                                 /*return_pc_adjustment*/ 0,
-                                                                 RegisterSaver::return_pc_is_lr);
+  map = RegisterSaver::push_frame_reg_args_and_save_live_registers(masm,
+                                                                   &frame_size_in_bytes,
+                                                                   /*generate_oop_map*/ true,
+                                                                   /*return_pc_adjustment*/ 0,
+                                                                   RegisterSaver::return_pc_is_lr);
 
   // Use noreg as last_Java_pc, the return pc will be reconstructed
   // from the physical frame.

src/cpu/ppc/vm/stubGenerator_ppc.cpp

@@ -79,11 +79,11 @@ class StubGenerator: public StubCodeGenerator {
 
     StubCodeMark mark(this, "StubRoutines", "call_stub");
 
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
     // some sanity checks
-    assert((sizeof(frame::abi_48) % 16) == 0,                 "unaligned");
-    assert((sizeof(frame::abi_112) % 16) == 0,                "unaligned");
+    assert((sizeof(frame::abi_minframe) % 16) == 0,           "unaligned");
+    assert((sizeof(frame::abi_reg_args) % 16) == 0,           "unaligned");
     assert((sizeof(frame::spill_nonvolatiles) % 16) == 0,     "unaligned");
     assert((sizeof(frame::parent_ijava_frame_abi) % 16) == 0, "unaligned");
     assert((sizeof(frame::entry_frame_locals) % 16) == 0,     "unaligned");
@@ -444,7 +444,7 @@ class StubGenerator: public StubCodeGenerator {
 
     // Save LR/CR and copy exception pc (LR) into R4_ARG2.
     __ save_LR_CR(R4_ARG2);
-    __ push_frame_abi112(0, R0);
+    __ push_frame_reg_args(0, R0);
     // Find exception handler.
     __ call_VM_leaf(CAST_FROM_FN_PTR(address,
                      SharedRuntime::exception_handler_for_return_address),
@@ -519,7 +519,7 @@ class StubGenerator: public StubCodeGenerator {
     MacroAssembler* masm = new MacroAssembler(&code);
 
     OopMapSet* oop_maps  = new OopMapSet();
-    int frame_size_in_bytes = frame::abi_112_size;
+    int frame_size_in_bytes = frame::abi_reg_args_size;
     OopMap* map = new OopMap(frame_size_in_bytes / sizeof(jint), 0);
 
     StubCodeMark mark(this, "StubRoutines", "throw_exception");
@@ -529,7 +529,7 @@ class StubGenerator: public StubCodeGenerator {
     __ save_LR_CR(R11_scratch1);
 
     // Push a frame.
-    __ push_frame_abi112(0, R11_scratch1);
+    __ push_frame_reg_args(0, R11_scratch1);
 
     address frame_complete_pc = __ pc();
 
@@ -551,8 +551,11 @@ class StubGenerator: public StubCodeGenerator {
     if (arg2 != noreg) {
       __ mr(R5_ARG3, arg2);
     }
-    __ call_c(CAST_FROM_FN_PTR(FunctionDescriptor*, runtime_entry),
-              relocInfo::none);
+#if defined(ABI_ELFv2)
+    __ call_c(runtime_entry, relocInfo::none);
+#else
+    __ call_c(CAST_FROM_FN_PTR(FunctionDescriptor*, runtime_entry), relocInfo::none);
+#endif
 
     // Set an oopmap for the call site.
     oop_maps->add_gc_map((int)(gc_map_pc - start), map);
@@ -614,7 +617,7 @@ class StubGenerator: public StubCodeGenerator {
         // With G1, don't generate the call if we statically know that the target in uninitialized
         if (!dest_uninitialized) {
           const int spill_slots = 4 * wordSize;
-          const int frame_size  = frame::abi_112_size + spill_slots;
+          const int frame_size  = frame::abi_reg_args_size + spill_slots;
           Label filtered;
 
           // Is marking active?
@@ -628,7 +631,7 @@ class StubGenerator: public StubCodeGenerator {
           __ beq(CCR0, filtered);
 
           __ save_LR_CR(R0);
-          __ push_frame_abi112(spill_slots, R0);
+          __ push_frame_reg_args(spill_slots, R0);
           __ std(from,  frame_size - 1 * wordSize, R1_SP);
           __ std(to,    frame_size - 2 * wordSize, R1_SP);
           __ std(count, frame_size - 3 * wordSize, R1_SP);
@@ -672,7 +675,7 @@ class StubGenerator: public StubCodeGenerator {
           if (branchToEnd) {
             __ save_LR_CR(R0);
             // We need this frame only to spill LR.
-            __ push_frame_abi112(0, R0);
+            __ push_frame_reg_args(0, R0);
             __ call_VM_leaf(CAST_FROM_FN_PTR(address, BarrierSet::static_write_ref_array_post), addr, count);
             __ pop_frame();
             __ restore_LR_CR(R0);
@@ -742,7 +745,7 @@ class StubGenerator: public StubCodeGenerator {
     StubCodeMark mark(this, "StubRoutines", "zero_words_aligned8");
 
     // Implemented as in ClearArray.
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
     Register base_ptr_reg   = R3_ARG1; // tohw (needs to be 8b aligned)
     Register cnt_dwords_reg = R4_ARG2; // count (in dwords)
@@ -820,7 +823,7 @@ class StubGenerator: public StubCodeGenerator {
   //
   address generate_handler_for_unsafe_access() {
     StubCodeMark mark(this, "StubRoutines", "handler_for_unsafe_access");
-    address start = __ emit_fd();
+    address start = __ function_entry();
     __ unimplemented("StubRoutines::handler_for_unsafe_access", 93);
     return start;
   }
@@ -861,7 +864,7 @@ class StubGenerator: public StubCodeGenerator {
   // to read from the safepoint polling page.
   address generate_load_from_poll() {
     StubCodeMark mark(this, "StubRoutines", "generate_load_from_poll");
-    address start = __ emit_fd();
+    address start = __ function_entry();
     __ unimplemented("StubRoutines::verify_oop", 95);  // TODO PPC port
     return start;
   }
@@ -885,7 +888,7 @@ class StubGenerator: public StubCodeGenerator {
   //
   address generate_fill(BasicType t, bool aligned, const char* name) {
     StubCodeMark mark(this, "StubRoutines", name);
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
     const Register to    = R3_ARG1;   // source array address
     const Register value = R4_ARG2;   // fill value
@@ -1123,7 +1126,7 @@ class StubGenerator: public StubCodeGenerator {
   //
   address generate_disjoint_byte_copy(bool aligned, const char * name) {
     StubCodeMark mark(this, "StubRoutines", name);
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
     Register tmp1 = R6_ARG4;
     Register tmp2 = R7_ARG5;
@@ -1254,15 +1257,21 @@ class StubGenerator: public StubCodeGenerator {
   //
   address generate_conjoint_byte_copy(bool aligned, const char * name) {
     StubCodeMark mark(this, "StubRoutines", name);
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
     Register tmp1 = R6_ARG4;
     Register tmp2 = R7_ARG5;
     Register tmp3 = R8_ARG6;
 
+#if defined(ABI_ELFv2)
+     address nooverlap_target = aligned ?
+       StubRoutines::arrayof_jbyte_disjoint_arraycopy() :
+       StubRoutines::jbyte_disjoint_arraycopy();
+#else
     address nooverlap_target = aligned ?
       ((FunctionDescriptor*)StubRoutines::arrayof_jbyte_disjoint_arraycopy())->entry() :
       ((FunctionDescriptor*)StubRoutines::jbyte_disjoint_arraycopy())->entry();
+#endif
 
     array_overlap_test(nooverlap_target, 0);
     // Do reverse copy. We assume the case of actual overlap is rare enough
@@ -1345,7 +1354,7 @@ class StubGenerator: public StubCodeGenerator {
     Register tmp3 = R8_ARG6;
     Register tmp4 = R9_ARG7;
 
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
       Label l_1, l_2, l_3, l_4, l_5, l_6, l_7, l_8;
     // don't try anything fancy if arrays don't have many elements
@@ -1474,15 +1483,21 @@ class StubGenerator: public StubCodeGenerator {
   //
   address generate_conjoint_short_copy(bool aligned, const char * name) {
     StubCodeMark mark(this, "StubRoutines", name);
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
     Register tmp1 = R6_ARG4;
     Register tmp2 = R7_ARG5;
     Register tmp3 = R8_ARG6;
 
+#if defined(ABI_ELFv2)
+    address nooverlap_target = aligned ?
+        StubRoutines::arrayof_jshort_disjoint_arraycopy() :
+        StubRoutines::jshort_disjoint_arraycopy();
+#else
     address nooverlap_target = aligned ?
         ((FunctionDescriptor*)StubRoutines::arrayof_jshort_disjoint_arraycopy())->entry() :
         ((FunctionDescriptor*)StubRoutines::jshort_disjoint_arraycopy())->entry();
+#endif
 
     array_overlap_test(nooverlap_target, 1);
 
@@ -1597,7 +1612,7 @@ class StubGenerator: public StubCodeGenerator {
   //
   address generate_disjoint_int_copy(bool aligned, const char * name) {
     StubCodeMark mark(this, "StubRoutines", name);
-    address start = __ emit_fd();
+    address start = __ function_entry();
     generate_disjoint_int_copy_core(aligned);
     __ blr();
     return start;
@@ -1681,11 +1696,17 @@ class StubGenerator: public StubCodeGenerator {
   //
   address generate_conjoint_int_copy(bool aligned, const char * name) {
     StubCodeMark mark(this, "StubRoutines", name);
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
+#if defined(ABI_ELFv2)
+    address nooverlap_target = aligned ?
+      StubRoutines::arrayof_jint_disjoint_arraycopy() :
+      StubRoutines::jint_disjoint_arraycopy();
+#else
     address nooverlap_target = aligned ?
       ((FunctionDescriptor*)StubRoutines::arrayof_jint_disjoint_arraycopy())->entry() :
       ((FunctionDescriptor*)StubRoutines::jint_disjoint_arraycopy())->entry();
+#endif
 
     array_overlap_test(nooverlap_target, 2);
 
@@ -1767,7 +1788,7 @@ class StubGenerator: public StubCodeGenerator {
   //
   address generate_disjoint_long_copy(bool aligned, const char * name) {
     StubCodeMark mark(this, "StubRoutines", name);
-    address start = __ emit_fd();
+    address start = __ function_entry();
     generate_disjoint_long_copy_core(aligned);
     __ blr();
 
@@ -1849,11 +1870,17 @@ class StubGenerator: public StubCodeGenerator {
   //
   address generate_conjoint_long_copy(bool aligned, const char * name) {
     StubCodeMark mark(this, "StubRoutines", name);
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
+#if defined(ABI_ELFv2)
+    address nooverlap_target = aligned ?
+      StubRoutines::arrayof_jlong_disjoint_arraycopy() :
+      StubRoutines::jlong_disjoint_arraycopy();
+#else
     address nooverlap_target = aligned ?
       ((FunctionDescriptor*)StubRoutines::arrayof_jlong_disjoint_arraycopy())->entry() :
       ((FunctionDescriptor*)StubRoutines::jlong_disjoint_arraycopy())->entry();
+#endif
 
     array_overlap_test(nooverlap_target, 3);
     generate_conjoint_long_copy_core(aligned);
@@ -1875,11 +1902,17 @@ class StubGenerator: public StubCodeGenerator {
   address generate_conjoint_oop_copy(bool aligned, const char * name, bool dest_uninitialized) {
     StubCodeMark mark(this, "StubRoutines", name);
 
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
+#if defined(ABI_ELFv2)
+    address nooverlap_target = aligned ?
+      StubRoutines::arrayof_oop_disjoint_arraycopy() :
+      StubRoutines::oop_disjoint_arraycopy();
+#else
     address nooverlap_target = aligned ?
       ((FunctionDescriptor*)StubRoutines::arrayof_oop_disjoint_arraycopy())->entry() :
       ((FunctionDescriptor*)StubRoutines::oop_disjoint_arraycopy())->entry();
+#endif
 
     gen_write_ref_array_pre_barrier(R3_ARG1, R4_ARG2, R5_ARG3, dest_uninitialized, R9_ARG7);
 
@@ -1910,7 +1943,7 @@ class StubGenerator: public StubCodeGenerator {
   //
   address generate_disjoint_oop_copy(bool aligned, const char * name, bool dest_uninitialized) {
     StubCodeMark mark(this, "StubRoutines", name);
-    address start = __ emit_fd();
+    address start = __ function_entry();
 
     gen_write_ref_array_pre_barrier(R3_ARG1, R4_ARG2, R5_ARG3, dest_uninitialized, R9_ARG7);
 
@@ -1991,7 +2024,7 @@ class StubGenerator: public StubCodeGenerator {
     StubCodeMark mark(this, "StubRoutines", name);
 
     // Entry point, pc or function descriptor.
-    *entry = __ emit_fd();
+    *entry = __ function_entry();
 
     // Load *adr into R4_ARG2, may fault.
     *fault_pc = __ pc();

src/cpu/ppc/vm/vm_version_ppc.cpp

@@ -24,7 +24,8 @@
  */
 
 #include "precompiled.hpp"
-#include "assembler_ppc.inline.hpp"
+#include "asm/assembler.inline.hpp"
+#include "asm/macroAssembler.inline.hpp"
 #include "compiler/disassembler.hpp"
 #include "memory/resourceArea.hpp"
 #include "runtime/java.hpp"
@@ -168,7 +169,7 @@ void VM_Version::determine_section_size() {
 
   uint32_t *code = (uint32_t *)a->pc();
   // Emit code.
-  void (*test1)() = (void(*)())(void *)a->emit_fd();
+  void (*test1)() = (void(*)())(void *)a->function_entry();
 
   Label l1;
 
@@ -242,7 +243,7 @@ void VM_Version::determine_section_size() {
   a->blr();
 
   // Emit code.
-  void (*test2)() = (void(*)())(void *)a->emit_fd();
+  void (*test2)() = (void(*)())(void *)a->function_entry();
   // uint32_t *code = (uint32_t *)a->pc();
 
   Label l2;
@@ -383,8 +384,12 @@ void VM_Version::determine_section_size() {
 #endif // COMPILER2
 
 void VM_Version::determine_features() {
+#if defined(ABI_ELFv2)
+  const int code_size = (num_features+1+2*7)*BytesPerInstWord; // TODO(asmundak): calculation is incorrect.
+#else
   // 7 InstWords for each call (function descriptor + blr instruction).
   const int code_size = (num_features+1+2*7)*BytesPerInstWord;
+#endif
   int features = 0;
 
   // create test area
@@ -398,7 +403,7 @@ void VM_Version::determine_features() {
   MacroAssembler* a = new MacroAssembler(&cb);
 
   // Emit code.
-  void (*test)(address addr, uint64_t offset)=(void(*)(address addr, uint64_t offset))(void *)a->emit_fd();
+  void (*test)(address addr, uint64_t offset)=(void(*)(address addr, uint64_t offset))(void *)a->function_entry();
   uint32_t *code = (uint32_t *)a->pc();
   // Don't use R0 in ldarx.
   // Keep R3_ARG1 unmodified, it contains &field (see below).
@@ -415,7 +420,7 @@ void VM_Version::determine_features() {
   a->blr();
 
   // Emit function to set one cache line to zero. Emit function descriptor and get pointer to it.
-  void (*zero_cacheline_func_ptr)(char*) = (void(*)(char*))(void *)a->emit_fd();
+  void (*zero_cacheline_func_ptr)(char*) = (void(*)(char*))(void *)a->function_entry();
   a->dcbz(R3_ARG1); // R3_ARG1 = addr
   a->blr();
 

src/share/vm/utilities/elfFile.cpp

@@ -140,7 +140,7 @@ bool ElfFile::load_tables() {
       }
     }
 
-#if defined(PPC64)
+#if defined(PPC64) && !defined(ABI_ELFv2)
     // Now read the .opd section wich contains the PPC64 function descriptor table.
     // The .opd section is only available on PPC64 (see for example:
     // http://refspecs.linuxfoundation.org/LSB_3.1.1/LSB-Core-PPC64/LSB-Core-PPC64/specialsections.html)