/*

 * Copyright (c) 1997, 2018, 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.

 *

 */

#include "precompiled.hpp"

#include "interp_masm_x86.hpp"

#include "interpreter/interpreter.hpp"

#include "interpreter/interpreterRuntime.hpp"

#include "logging/log.hpp"

#include "oops/arrayOop.hpp"

#include "oops/markOop.hpp"

#include "oops/methodData.hpp"

#include "oops/method.hpp"

#include "prims/jvmtiExport.hpp"

#include "prims/jvmtiThreadState.hpp"

#include "runtime/basicLock.hpp"

#include "runtime/biasedLocking.hpp"

#include "runtime/frame.inline.hpp"

#include "runtime/safepointMechanism.hpp"

#include "runtime/sharedRuntime.hpp"

#include "runtime/thread.inline.hpp"

// Implementation of InterpreterMacroAssembler

void InterpreterMacroAssembler::jump_to_entry(address entry) {

  assert(entry, "Entry must have been generated by now");

  jump(RuntimeAddress(entry));

}

void InterpreterMacroAssembler::profile_obj_type(Register obj, const Address& mdo_addr) {

  Label update, next, none;

  verify_oop(obj);

  testptr(obj, obj);

  jccb(Assembler::notZero, update);

  orptr(mdo_addr, TypeEntries::null_seen);

  jmpb(next);

  bind(update);

  load_klass(obj, obj);

  xorptr(obj, mdo_addr);

  testptr(obj, TypeEntries::type_klass_mask);

  jccb(Assembler::zero, next); // klass seen before, nothing to

                               // do. The unknown bit may have been

                               // set already but no need to check.

  testptr(obj, TypeEntries::type_unknown);

  jccb(Assembler::notZero, next); // already unknown. Nothing to do anymore.

  cmpptr(mdo_addr, 0);

  jccb(Assembler::equal, none);

  cmpptr(mdo_addr, TypeEntries::null_seen);

  jccb(Assembler::equal, none);

  // There is a chance that the checks above (re-reading profiling

  // data from memory) fail if another thread has just set the

  // profiling to this obj's klass

  xorptr(obj, mdo_addr);

  testptr(obj, TypeEntries::type_klass_mask);

  jccb(Assembler::zero, next);

  // different than before. Cannot keep accurate profile.

  orptr(mdo_addr, TypeEntries::type_unknown);

  jmpb(next);

  bind(none);

  // first time here. Set profile type.

  movptr(mdo_addr, obj);

  bind(next);

}

void InterpreterMacroAssembler::profile_arguments_type(Register mdp, Register callee, Register tmp, bool is_virtual) {

  if (!ProfileInterpreter) {

    return;

  }

  if (MethodData::profile_arguments() || MethodData::profile_return()) {

    Label profile_continue;

    test_method_data_pointer(mdp, profile_continue);

    int off_to_start = is_virtual ? in_bytes(VirtualCallData::virtual_call_data_size()) : in_bytes(CounterData::counter_data_size());

    cmpb(Address(mdp, in_bytes(DataLayout::tag_offset()) - off_to_start), is_virtual ? DataLayout::virtual_call_type_data_tag : DataLayout::call_type_data_tag);

    jcc(Assembler::notEqual, profile_continue);

    if (MethodData::profile_arguments()) {

      Label done;

      int off_to_args = in_bytes(TypeEntriesAtCall::args_data_offset());

      addptr(mdp, off_to_args);

      for (int i = 0; i < TypeProfileArgsLimit; i++) {

        if (i > 0 || MethodData::profile_return()) {

          // If return value type is profiled we may have no argument to profile

          movptr(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::cell_count_offset())-off_to_args));

          subl(tmp, i*TypeStackSlotEntries::per_arg_count());

          cmpl(tmp, TypeStackSlotEntries::per_arg_count());

          jcc(Assembler::less, done);

        }

        movptr(tmp, Address(callee, Method::const_offset()));

        load_unsigned_short(tmp, Address(tmp, ConstMethod::size_of_parameters_offset()));

        // stack offset o (zero based) from the start of the argument

        // list, for n arguments translates into offset n - o - 1 from

        // the end of the argument list

        subptr(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::stack_slot_offset(i))-off_to_args));

        subl(tmp, 1);

        Address arg_addr = argument_address(tmp);

        movptr(tmp, arg_addr);

        Address mdo_arg_addr(mdp, in_bytes(TypeEntriesAtCall::argument_type_offset(i))-off_to_args);

        profile_obj_type(tmp, mdo_arg_addr);

        int to_add = in_bytes(TypeStackSlotEntries::per_arg_size());

        addptr(mdp, to_add);

        off_to_args += to_add;

      }

      if (MethodData::profile_return()) {

        movptr(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::cell_count_offset())-off_to_args));

        subl(tmp, TypeProfileArgsLimit*TypeStackSlotEntries::per_arg_count());

      }

      bind(done);

      if (MethodData::profile_return()) {

        // We're right after the type profile for the last

        // argument. tmp is the number of cells left in the

        // CallTypeData/VirtualCallTypeData to reach its end. Non null

        // if there's a return to profile.

        assert(ReturnTypeEntry::static_cell_count() < TypeStackSlotEntries::per_arg_count(), "can't move past ret type");

        shll(tmp, exact_log2(DataLayout::cell_size));

        addptr(mdp, tmp);

      }

      movptr(Address(rbp, frame::interpreter_frame_mdp_offset * wordSize), mdp);

    } else {

      assert(MethodData::profile_return(), "either profile call args or call ret");

      update_mdp_by_constant(mdp, in_bytes(TypeEntriesAtCall::return_only_size()));

    }

    // mdp points right after the end of the

    // CallTypeData/VirtualCallTypeData, right after the cells for the

    // return value type if there's one

    bind(profile_continue);

  }

}

void InterpreterMacroAssembler::profile_return_type(Register mdp, Register ret, Register tmp) {

  assert_different_registers(mdp, ret, tmp, _bcp_register);

  if (ProfileInterpreter && MethodData::profile_return()) {

    Label profile_continue, done;

    test_method_data_pointer(mdp, profile_continue);

    if (MethodData::profile_return_jsr292_only()) {

      assert(Method::intrinsic_id_size_in_bytes() == 2, "assuming Method::_intrinsic_id is u2");

      // If we don't profile all invoke bytecodes we must make sure

      // it's a bytecode we indeed profile. We can't go back to the

      // begining of the ProfileData we intend to update to check its

      // type because we're right after it and we don't known its

      // length

      Label do_profile;

      cmpb(Address(_bcp_register, 0), Bytecodes::_invokedynamic);

      jcc(Assembler::equal, do_profile);

      cmpb(Address(_bcp_register, 0), Bytecodes::_invokehandle);

      jcc(Assembler::equal, do_profile);

      get_method(tmp);

      cmpw(Address(tmp, Method::intrinsic_id_offset_in_bytes()), vmIntrinsics::_compiledLambdaForm);

      jcc(Assembler::notEqual, profile_continue);

      bind(do_profile);

    }

    Address mdo_ret_addr(mdp, -in_bytes(ReturnTypeEntry::size()));

    mov(tmp, ret);

    profile_obj_type(tmp, mdo_ret_addr);

    bind(profile_continue);

  }

}

void InterpreterMacroAssembler::profile_parameters_type(Register mdp, Register tmp1, Register tmp2) {

  if (ProfileInterpreter && MethodData::profile_parameters()) {

    Label profile_continue, done;

    test_method_data_pointer(mdp, profile_continue);

    // Load the offset of the area within the MDO used for

    // parameters. If it's negative we're not profiling any parameters

    movl(tmp1, Address(mdp, in_bytes(MethodData::parameters_type_data_di_offset()) - in_bytes(MethodData::data_offset())));

    testl(tmp1, tmp1);

    jcc(Assembler::negative, profile_continue);

    // Compute a pointer to the area for parameters from the offset

    // and move the pointer to the slot for the last

    // parameters. Collect profiling from last parameter down.

    // mdo start + parameters offset + array length - 1

    addptr(mdp, tmp1);

    movptr(tmp1, Address(mdp, ArrayData::array_len_offset()));

    decrement(tmp1, TypeStackSlotEntries::per_arg_count());

    Label loop;

    bind(loop);

    int off_base = in_bytes(ParametersTypeData::stack_slot_offset(0));

    int type_base = in_bytes(ParametersTypeData::type_offset(0));

    Address::ScaleFactor per_arg_scale = Address::times(DataLayout::cell_size);

    Address arg_off(mdp, tmp1, per_arg_scale, off_base);

    Address arg_type(mdp, tmp1, per_arg_scale, type_base);

    // load offset on the stack from the slot for this parameter

    movptr(tmp2, arg_off);

    negptr(tmp2);

    // read the parameter from the local area

    movptr(tmp2, Address(_locals_register, tmp2, Interpreter::stackElementScale()));

    // profile the parameter

    profile_obj_type(tmp2, arg_type);

    // go to next parameter

    decrement(tmp1, TypeStackSlotEntries::per_arg_count());

    jcc(Assembler::positive, loop);

    bind(profile_continue);

  }

}

void InterpreterMacroAssembler::call_VM_leaf_base(address entry_point,

                                                  int number_of_arguments) {

  // interpreter specific

  //

  // Note: No need to save/restore bcp & locals registers

  //       since these are callee saved registers and no blocking/

  //       GC can happen in leaf calls.

  // Further Note: DO NOT save/restore bcp/locals. If a caller has

  // already saved them so that it can use rsi/rdi as temporaries

  // then a save/restore here will DESTROY the copy the caller

  // saved! There used to be a save_bcp() that only happened in

  // the ASSERT path (no restore_bcp). Which caused bizarre failures

  // when jvm built with ASSERTs.

#ifdef ASSERT

  {

    Label L;

    cmpptr(Address(rbp, frame::interpreter_frame_last_sp_offset * wordSize), (int32_t)NULL_WORD);

    jcc(Assembler::equal, L);

    stop("InterpreterMacroAssembler::call_VM_leaf_base:"

         " last_sp != NULL");

    bind(L);

  }

#endif

  // super call

  MacroAssembler::call_VM_leaf_base(entry_point, number_of_arguments);

  // interpreter specific

  // LP64: Used to ASSERT that r13/r14 were equal to frame's bcp/locals

  // but since they may not have been saved (and we don't want to

  // save them here (see note above) the assert is invalid.

}

void InterpreterMacroAssembler::call_VM_base(Register oop_result,

                                             Register java_thread,

                                             Register last_java_sp,

                                             address  entry_point,

                                             int      number_of_arguments,

                                             bool     check_exceptions) {

  // interpreter specific

  //

  // Note: Could avoid restoring locals ptr (callee saved) - however doesn't

  //       really make a difference for these runtime calls, since they are

  //       slow anyway. Btw., bcp must be saved/restored since it may change

  //       due to GC.

  NOT_LP64(assert(java_thread == noreg , "not expecting a precomputed java thread");)

  save_bcp();

#ifdef ASSERT

  {

    Label L;

    cmpptr(Address(rbp, frame::interpreter_frame_last_sp_offset * wordSize), (int32_t)NULL_WORD);

    jcc(Assembler::equal, L);

    stop("InterpreterMacroAssembler::call_VM_base:"

         " last_sp != NULL");

    bind(L);

  }

#endif /* ASSERT */

  // super call

  MacroAssembler::call_VM_base(oop_result, noreg, last_java_sp,

                               entry_point, number_of_arguments,

                               check_exceptions);

  // interpreter specific

  restore_bcp();

  restore_locals();

}

void InterpreterMacroAssembler::check_and_handle_popframe(Register java_thread) {

  if (JvmtiExport::can_pop_frame()) {

    Label L;

    // Initiate popframe handling only if it is not already being

    // processed.  If the flag has the popframe_processing bit set, it

    // means that this code is called *during* popframe handling - we

    // don't want to reenter.

    // This method is only called just after the call into the vm in

    // call_VM_base, so the arg registers are available.

    Register pop_cond = NOT_LP64(java_thread) // Not clear if any other register is available on 32 bit

                        LP64_ONLY(c_rarg0);

    movl(pop_cond, Address(java_thread, JavaThread::popframe_condition_offset()));

    testl(pop_cond, JavaThread::popframe_pending_bit);

    jcc(Assembler::zero, L);

    testl(pop_cond, JavaThread::popframe_processing_bit);

    jcc(Assembler::notZero, L);

    // Call Interpreter::remove_activation_preserving_args_entry() to get the

    // address of the same-named entrypoint in the generated interpreter code.

    call_VM_leaf(CAST_FROM_FN_PTR(address, Interpreter::remove_activation_preserving_args_entry));

    jmp(rax);

    bind(L);

    NOT_LP64(get_thread(java_thread);)

  }

}

void InterpreterMacroAssembler::load_earlyret_value(TosState state) {

  Register thread = LP64_ONLY(r15_thread) NOT_LP64(rcx);

  NOT_LP64(get_thread(thread);)

  movptr(rcx, Address(thread, JavaThread::jvmti_thread_state_offset()));

  const Address tos_addr(rcx, JvmtiThreadState::earlyret_tos_offset());

  const Address oop_addr(rcx, JvmtiThreadState::earlyret_oop_offset());

  const Address val_addr(rcx, JvmtiThreadState::earlyret_value_offset());

#ifdef _LP64

  switch (state) {

    case atos: movptr(rax, oop_addr);

               movptr(oop_addr, (int32_t)NULL_WORD);

               verify_oop(rax, state);              break;

    case ltos: movptr(rax, val_addr);                 break;

    case btos:                                   // fall through

    case ztos:                                   // fall through

    case ctos:                                   // fall through

    case stos:                                   // fall through

    case itos: movl(rax, val_addr);                 break;

    case ftos: load_float(val_addr);                break;

    case dtos: load_double(val_addr);               break;

    case vtos: /* nothing to do */                  break;

    default  : ShouldNotReachHere();

  }

  // Clean up tos value in the thread object

  movl(tos_addr,  (int) ilgl);

  movl(val_addr,  (int32_t) NULL_WORD);

#else

  const Address val_addr1(rcx, JvmtiThreadState::earlyret_value_offset()

                             + in_ByteSize(wordSize));

  switch (state) {

    case atos: movptr(rax, oop_addr);

               movptr(oop_addr, NULL_WORD);

               verify_oop(rax, state);                break;

    case ltos:

               movl(rdx, val_addr1);               // fall through

    case btos:                                     // fall through

    case ztos:                                     // fall through

    case ctos:                                     // fall through

    case stos:                                     // fall through

    case itos: movl(rax, val_addr);                   break;

    case ftos: load_float(val_addr);                  break;

    case dtos: load_double(val_addr);                 break;

    case vtos: /* nothing to do */                    break;

    default  : ShouldNotReachHere();

  }

#endif // _LP64

  // Clean up tos value in the thread object

  movl(tos_addr,  (int32_t) ilgl);

  movptr(val_addr,  NULL_WORD);

  NOT_LP64(movptr(val_addr1, NULL_WORD);)

}

void InterpreterMacroAssembler::check_and_handle_earlyret(Register java_thread) {

  if (JvmtiExport::can_force_early_return()) {

    Label L;

    Register tmp = LP64_ONLY(c_rarg0) NOT_LP64(java_thread);

    Register rthread = LP64_ONLY(r15_thread) NOT_LP64(java_thread);

    movptr(tmp, Address(rthread, JavaThread::jvmti_thread_state_offset()));

    testptr(tmp, tmp);

    jcc(Assembler::zero, L); // if (thread->jvmti_thread_state() == NULL) exit;

    // Initiate earlyret handling only if it is not already being processed.

    // If the flag has the earlyret_processing bit set, it means that this code

    // is called *during* earlyret handling - we don't want to reenter.

    movl(tmp, Address(tmp, JvmtiThreadState::earlyret_state_offset()));

    cmpl(tmp, JvmtiThreadState::earlyret_pending);

    jcc(Assembler::notEqual, L);

    // Call Interpreter::remove_activation_early_entry() to get the address of the

    // same-named entrypoint in the generated interpreter code.

    NOT_LP64(get_thread(java_thread);)

    movptr(tmp, Address(rthread, JavaThread::jvmti_thread_state_offset()));

#ifdef _LP64

    movl(tmp, Address(tmp, JvmtiThreadState::earlyret_tos_offset()));

    call_VM_leaf(CAST_FROM_FN_PTR(address, Interpreter::remove_activation_early_entry), tmp);

#else

    pushl(Address(tmp, JvmtiThreadState::earlyret_tos_offset()));

    call_VM_leaf(CAST_FROM_FN_PTR(address, Interpreter::remove_activation_early_entry), 1);

#endif // _LP64

    jmp(rax);

    bind(L);

    NOT_LP64(get_thread(java_thread);)

  }

}

void InterpreterMacroAssembler::get_unsigned_2_byte_index_at_bcp(Register reg, int bcp_offset) {

  assert(bcp_offset >= 0, "bcp is still pointing to start of bytecode");

  load_unsigned_short(reg, Address(_bcp_register, bcp_offset));

  bswapl(reg);

  shrl(reg, 16);

}

void InterpreterMacroAssembler::get_cache_index_at_bcp(Register index,

                                                       int bcp_offset,

                                                       size_t index_size) {

  assert(bcp_offset > 0, "bcp is still pointing to start of bytecode");

  if (index_size == sizeof(u2)) {

    load_unsigned_short(index, Address(_bcp_register, bcp_offset));

  } else if (index_size == sizeof(u4)) {

    movl(index, Address(_bcp_register, bcp_offset));

    // Check if the secondary index definition is still ~x, otherwise

    // we have to change the following assembler code to calculate the

    // plain index.

    assert(ConstantPool::decode_invokedynamic_index(~123) == 123, "else change next line");

    notl(index);  // convert to plain index

  } else if (index_size == sizeof(u1)) {

    load_unsigned_byte(index, Address(_bcp_register, bcp_offset));

  } else {

    ShouldNotReachHere();

  }

}

void InterpreterMacroAssembler::get_cache_and_index_at_bcp(Register cache,

                                                           Register index,

                                                           int bcp_offset,

                                                           size_t index_size) {

  assert_different_registers(cache, index);

  get_cache_index_at_bcp(index, bcp_offset, index_size);

  movptr(cache, Address(rbp, frame::interpreter_frame_cache_offset * wordSize));

  assert(sizeof(ConstantPoolCacheEntry) == 4 * wordSize, "adjust code below");

  // convert from field index to ConstantPoolCacheEntry index

  assert(exact_log2(in_words(ConstantPoolCacheEntry::size())) == 2, "else change next line");

  shll(index, 2);

}

void InterpreterMacroAssembler::get_cache_and_index_and_bytecode_at_bcp(Register cache,

                                                                        Register index,

                                                                        Register bytecode,

                                                                        int byte_no,

                                                                        int bcp_offset,

                                                                        size_t index_size) {

  get_cache_and_index_at_bcp(cache, index, bcp_offset, index_size);

  // We use a 32-bit load here since the layout of 64-bit words on

  // little-endian machines allow us that.

  movl(bytecode, Address(cache, index, Address::times_ptr, ConstantPoolCache::base_offset() + ConstantPoolCacheEntry::indices_offset()));

  const int shift_count = (1 + byte_no) * BitsPerByte;

  assert((byte_no == TemplateTable::f1_byte && shift_count == ConstantPoolCacheEntry::bytecode_1_shift) ||

         (byte_no == TemplateTable::f2_byte && shift_count == ConstantPoolCacheEntry::bytecode_2_shift),

         "correct shift count");

  shrl(bytecode, shift_count);

  assert(ConstantPoolCacheEntry::bytecode_1_mask == ConstantPoolCacheEntry::bytecode_2_mask, "common mask");

  andl(bytecode, ConstantPoolCacheEntry::bytecode_1_mask);

}

void InterpreterMacroAssembler::get_cache_entry_pointer_at_bcp(Register cache,

                                                               Register tmp,

                                                               int bcp_offset,

                                                               size_t index_size) {

  assert(cache != tmp, "must use different register");

  get_cache_index_at_bcp(tmp, bcp_offset, index_size);

  assert(sizeof(ConstantPoolCacheEntry) == 4 * wordSize, "adjust code below");

  // convert from field index to ConstantPoolCacheEntry index

  // and from word offset to byte offset

  assert(exact_log2(in_bytes(ConstantPoolCacheEntry::size_in_bytes())) == 2 + LogBytesPerWord, "else change next line");

  shll(tmp, 2 + LogBytesPerWord);

  movptr(cache, Address(rbp, frame::interpreter_frame_cache_offset * wordSize));

  // skip past the header

  addptr(cache, in_bytes(ConstantPoolCache::base_offset()));

  addptr(cache, tmp);  // construct pointer to cache entry

}

// Load object from cpool->resolved_references(index)

void InterpreterMacroAssembler::load_resolved_reference_at_index(

                                           Register result, Register index, Register tmp) {

  assert_different_registers(result, index);