/*
 * Copyright (c) 2016, 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 "asm/macroAssembler.inline.hpp"
#include "c1/c1_LIRAssembler.hpp"
#include "c1/c1_MacroAssembler.hpp"
#include "gc/g1/c1G1BSCodeGen.hpp"
#include "gc/g1/g1CardTable.hpp"
#include "gc/g1/g1BarrierSet.hpp"
#include "gc/g1/g1BSCodeGen.hpp"
#include "gc/g1/heapRegion.hpp"
#include "gc/shared/collectedHeap.hpp"
#include "runtime/thread.hpp"
#include "interpreter/interp_masm.hpp"

#define __ masm->

void G1BSCodeGen::gen_write_ref_array_pre_barrier(MacroAssembler* masm, DecoratorSet decorators, Register addr, Register count) {
  bool dest_uninitialized = (decorators & DEST_NOT_INITIALIZED) != 0;

  if (!dest_uninitialized) {
     __ pusha();                      // push registers
     if (count == c_rarg0) {
       if (addr == c_rarg1) {
         // exactly backwards!!
         __ xchgptr(c_rarg1, c_rarg0);
       } else {
         __ movptr(c_rarg1, count);
         __ movptr(c_rarg0, addr);
       }
     } else {
       __ movptr(c_rarg0, addr);
       __ movptr(c_rarg1, count);
     }
     __ call_VM_leaf(CAST_FROM_FN_PTR(address, ModRefBarrierSet::static_write_ref_array_pre), 2);
     __ popa();
  }
}

void G1BSCodeGen::load_at(MacroAssembler* masm, DecoratorSet decorators, BasicType type,
                          Register dst, Address src, Register tmp1, Register tmp2) {
  bool on_oop = type == T_OBJECT || type == T_ARRAY;
  bool on_weak = (decorators & GC_ACCESS_ON_WEAK) != 0;
  bool on_phantom = (decorators & GC_ACCESS_ON_PHANTOM) != 0;
  bool on_reference = on_weak || on_phantom;
  ModRefBSCodeGen::load_at(masm, decorators, type, dst, src, tmp1, tmp2);
  if (on_oop && on_reference) {
    const Register thread = NOT_LP64(rcx) LP64_ONLY(r15_thread);
    NOT_LP64(__ get_thread(thread));

    // Generate the G1 pre-barrier code to log the value of
    // the referent field or jniHandle in an SATB buffer.
    g1_write_barrier_pre(masm /* masm */,
                         noreg /* obj */,
                         dst /* pre_val */,
                         thread /* thread */,
                         tmp1 /* tmp */,
                         true /* tosca_live */,
                         true /* expand_call */);
  }
}

void G1BSCodeGen::g1_write_barrier_pre(MacroAssembler* masm,
                                       Register obj,
                                       Register pre_val,
                                       Register thread,
                                       Register tmp,
                                       bool tosca_live,
                                       bool expand_call) {
  // If expand_call is true then we expand the call_VM_leaf macro
  // directly to skip generating the check by
  // InterpreterMacroAssembler::call_VM_leaf_base that checks _last_sp.

#ifdef _LP64
  assert(thread == r15_thread, "must be");
#endif // _LP64

  Label done;
  Label runtime;

  assert(pre_val != noreg, "check this code");

  if (obj != noreg) {
    assert_different_registers(obj, pre_val, tmp);
    assert(pre_val != rax, "check this code");
  }

  Address in_progress(thread, in_bytes(JavaThread::satb_mark_queue_offset() +
                                       SATBMarkQueue::byte_offset_of_active()));
  Address index(thread, in_bytes(JavaThread::satb_mark_queue_offset() +
                                       SATBMarkQueue::byte_offset_of_index()));
  Address buffer(thread, in_bytes(JavaThread::satb_mark_queue_offset() +
                                       SATBMarkQueue::byte_offset_of_buf()));


  // Is marking active?
  if (in_bytes(SATBMarkQueue::byte_width_of_active()) == 4) {
    __ cmpl(in_progress, 0);
  } else {
    assert(in_bytes(SATBMarkQueue::byte_width_of_active()) == 1, "Assumption");
    __ cmpb(in_progress, 0);
  }
  __ jcc(Assembler::equal, done);

  // Do we need to load the previous value?
  if (obj != noreg) {
    __ load_heap_oop(pre_val, Address(obj, 0));
  }

  // Is the previous value null?
  __ cmpptr(pre_val, (int32_t) NULL_WORD);
  __ jcc(Assembler::equal, done);

  // Can we store original value in the thread's buffer?
  // Is index == 0?
  // (The index field is typed as size_t.)

  __ movptr(tmp, index);                   // tmp := *index_adr
  __ cmpptr(tmp, 0);                       // tmp == 0?
  __ jcc(Assembler::equal, runtime);       // If yes, goto runtime

  __ subptr(tmp, wordSize);                // tmp := tmp - wordSize
  __ movptr(index, tmp);                   // *index_adr := tmp
  __ addptr(tmp, buffer);                  // tmp := tmp + *buffer_adr

  // Record the previous value
  __ movptr(Address(tmp, 0), pre_val);
  __ jmp(done);

  __ bind(runtime);
  // save the live input values
  if(tosca_live) __ push(rax);

  if (obj != noreg && obj != rax)
    __ push(obj);

  if (pre_val != rax)
    __ push(pre_val);

  // Calling the runtime using the regular call_VM_leaf mechanism generates
  // code (generated by InterpreterMacroAssember::call_VM_leaf_base)
  // that checks that the *(ebp+frame::interpreter_frame_last_sp) == NULL.
  //
  // If we care generating the pre-barrier without a frame (e.g. in the
  // intrinsified Reference.get() routine) then ebp might be pointing to
  // the caller frame and so this check will most likely fail at runtime.
  //
  // Expanding the call directly bypasses the generation of the check.
  // So when we do not have have a full interpreter frame on the stack
  // expand_call should be passed true.

  NOT_LP64( __ push(thread); )

  if (expand_call) {
    LP64_ONLY( assert(pre_val != c_rarg1, "smashed arg"); )
    if (c_rarg1 != thread) {
      __ mov(c_rarg1, thread);
    }
    if (c_rarg0 != pre_val) {
      __ mov(c_rarg0, pre_val);
    }
    __ MacroAssembler::call_VM_leaf_base(CAST_FROM_FN_PTR(address, G1BarrierSet::g1_wb_pre), 2);
  } else {
    __ call_VM_leaf(CAST_FROM_FN_PTR(address, G1BarrierSet::g1_wb_pre), pre_val, thread);
  }

  NOT_LP64( __ pop(thread); )

  // save the live input values
  if (pre_val != rax)
    __ pop(pre_val);

  if (obj != noreg && obj != rax)
    __ pop(obj);

  if(tosca_live) __ pop(rax);

  __ bind(done);
}

void G1BSCodeGen::g1_write_barrier_post(MacroAssembler* masm,
                                        Register store_addr,
                                        Register new_val,
                                        Register thread,
                                        Register tmp,
                                        Register tmp2) {
#ifdef _LP64
  assert(thread == r15_thread, "must be");
#endif // _LP64

  Address queue_index(thread, in_bytes(JavaThread::dirty_card_queue_offset() +
                                       DirtyCardQueue::byte_offset_of_index()));
  Address buffer(thread, in_bytes(JavaThread::dirty_card_queue_offset() +
                                       DirtyCardQueue::byte_offset_of_buf()));

  CardTableModRefBS* ct =
    barrier_set_cast<CardTableModRefBS>(Universe::heap()->barrier_set());
  assert(sizeof(*ct->card_table()->byte_map_base()) == sizeof(jbyte), "adjust this code");

  Label done;
  Label runtime;

  // Does store cross heap regions?

  __ movptr(tmp, store_addr);
  __ xorptr(tmp, new_val);
  __ shrptr(tmp, HeapRegion::LogOfHRGrainBytes);
  __ jcc(Assembler::equal, done);

  // crosses regions, storing NULL?

  __ cmpptr(new_val, (int32_t) NULL_WORD);
  __ jcc(Assembler::equal, done);

  // storing region crossing non-NULL, is card already dirty?

  const Register card_addr = tmp;
  const Register cardtable = tmp2;

  __ movptr(card_addr, store_addr);
  __ shrptr(card_addr, CardTable::card_shift);
  // Do not use ExternalAddress to load 'byte_map_base', since 'byte_map_base' is NOT
  // a valid address and therefore is not properly handled by the relocation code.
  __ movptr(cardtable, (intptr_t)ct->card_table()->byte_map_base());
  __ addptr(card_addr, cardtable);

  __ cmpb(Address(card_addr, 0), (int)G1CardTable::g1_young_card_val());
  __ jcc(Assembler::equal, done);

  __ membar(Assembler::Membar_mask_bits(Assembler::StoreLoad));
  __ cmpb(Address(card_addr, 0), (int)G1CardTable::dirty_card_val());
  __ jcc(Assembler::equal, done);


  // storing a region crossing, non-NULL oop, card is clean.
  // dirty card and log.

  __ movb(Address(card_addr, 0), (int)G1CardTable::dirty_card_val());

  __ cmpl(queue_index, 0);
  __ jcc(Assembler::equal, runtime);
  __ subl(queue_index, wordSize);
  __ movptr(tmp2, buffer);
#ifdef _LP64
  __ movslq(rscratch1, queue_index);
  __ addq(tmp2, rscratch1);
  __ movq(Address(tmp2, 0), card_addr);
#else
  __ addl(tmp2, queue_index);
  __ movl(Address(tmp2, 0), card_addr);
#endif
  __ jmp(done);

  __ bind(runtime);
  // save the live input values
  __ push(store_addr);
  __ push(new_val);
#ifdef _LP64
  __ call_VM_leaf(CAST_FROM_FN_PTR(address, G1BarrierSet::g1_wb_post), card_addr, r15_thread);
#else
  __ push(thread);
  __ call_VM_leaf(CAST_FROM_FN_PTR(address, G1BarrierSet::g1_wb_post), card_addr, thread);
  __ pop(thread);
#endif
  __ pop(new_val);
  __ pop(store_addr);

  __ bind(done);
}

void G1BSCodeGen::oop_store_at(MacroAssembler* masm, DecoratorSet decorators, BasicType type,
                               Address dst, Register val, Register tmp1, Register tmp2) {
  Register tmp3 = LP64_ONLY(r8) NOT_LP64(rsi);
  Register rthread = LP64_ONLY(r15_thread) NOT_LP64(rcx);
  // flatten object address if needed
  // We do it regardless of precise because we need the registers
  if (dst.index() == noreg && dst.disp() == 0) {
    if (dst.base() != tmp1) {
      __ movptr(tmp1, dst.base());
    }
  } else {
    __ lea(tmp1, dst);
  }

  NOT_LP64(__ get_thread(rcx));
  NOT_LP64(__ save_bcp());

  g1_write_barrier_pre(masm /*masm*/,
                       tmp1 /* obj */,
                       tmp2 /* pre_val */,
                       rthread /* thread */,
                       tmp3  /* tmp */,
                       val != noreg /* tosca_live */,
                       false /* expand_call */);
  if (val == noreg) {
    __ store_heap_oop_null(Address(tmp1, 0));
  } else {
    // G1 barrier needs uncompressed oop for region cross check.
    Register new_val = val;
    if (UseCompressedOops) {
      new_val = tmp2;
      __ movptr(new_val, val);
    }
    __ store_heap_oop(Address(tmp1, 0), val);
    g1_write_barrier_post(masm /*masm*/,
                          tmp1 /* store_adr */,
                          new_val /* new_val */,
                          rthread /* thread */,
                          tmp3 /* tmp */,
                          tmp2 /* tmp2 */);
  }
  NOT_LP64( __ restore_bcp());
}

void G1BSCodeGen::gen_write_ref_array_post_barrier(MacroAssembler* masm, DecoratorSet decorators, Register addr, Register count, Register tmp) {
  __ pusha();             // push registers (overkill)
  if (c_rarg0 == count) { // On win64 c_rarg0 == rcx
    assert_different_registers(c_rarg1, addr);
    __ mov(c_rarg1, count);
    __ mov(c_rarg0, addr);
  } else {
    assert_different_registers(c_rarg0, count);
    __ mov(c_rarg0, addr);
    __ mov(c_rarg1, count);
  }
  __ call_VM_leaf(CAST_FROM_FN_PTR(address, ModRefBarrierSet::static_write_ref_array_post), 2);
  __ popa();
}

#undef __
#define __ ce->masm()->

void G1BSCodeGen::gen_g1_pre_barrier_stub(LIR_Assembler* ce, G1PreBarrierStub* stub) {
  C1G1BSCodeGen* code_gen = (C1G1BSCodeGen*)Universe::heap()->barrier_set()->c1_code_gen();
  // At this point we know that marking is in progress.
  // If do_load() is true then we have to emit the
  // load of the previous value; otherwise it has already
  // been loaded into _pre_val.

  __ bind(stub->_entry);
  assert(stub->pre_val()->is_register(), "Precondition.");

  Register pre_val_reg = stub->pre_val()->as_register();

  if (stub->do_load()) {
    ce->mem2reg(stub->addr(), stub->pre_val(), T_OBJECT, stub->patch_code(), stub->info(), false /*wide*/, false /*unaligned*/);
  }

  __ cmpptr(pre_val_reg, (int32_t)NULL_WORD);
  __ jcc(Assembler::equal, stub->_continuation);
  ce->store_parameter(stub->pre_val()->as_register(), 0);
  __ call(RuntimeAddress(code_gen->pre_barrier_c1_runtime_code_blob()->code_begin()));
  __ jmp(stub->_continuation);

}

void G1BSCodeGen::gen_g1_post_barrier_stub(LIR_Assembler* ce, G1PostBarrierStub* stub) {
  C1G1BSCodeGen* code_gen = (C1G1BSCodeGen*)Universe::heap()->barrier_set()->c1_code_gen();
  __ bind(stub->_entry);
  assert(stub->addr()->is_register(), "Precondition.");
  assert(stub->new_val()->is_register(), "Precondition.");
  Register new_val_reg = stub->new_val()->as_register();
  __ cmpptr(new_val_reg, (int32_t) NULL_WORD);
  __ jcc(Assembler::equal, stub->_continuation);
  ce->store_parameter(stub->addr()->as_pointer_register(), 0);
  __ call(RuntimeAddress(code_gen->post_barrier_c1_runtime_code_blob()->code_begin()));
  __ jmp(stub->_continuation);
}

#undef __

#define __ sasm->

void G1BSCodeGen::generate_c1_pre_barrier_runtime_stub(StubAssembler* sasm) {
  __ prologue("g1_pre_barrier", false);
  // arg0 : previous value of memory

  BarrierSet* bs = Universe::heap()->barrier_set();
  __ push(rax);
  __ push(rdx);

  const Register pre_val = rax;
  const Register thread = NOT_LP64(rax) LP64_ONLY(r15_thread);
  const Register tmp = rdx;

  NOT_LP64(__ get_thread(thread);)

  Address queue_active(thread, in_bytes(JavaThread::satb_mark_queue_offset() +
                                        SATBMarkQueue::byte_offset_of_active()));
  Address queue_index(thread, in_bytes(JavaThread::satb_mark_queue_offset() +
                                       SATBMarkQueue::byte_offset_of_index()));
  Address buffer(thread, in_bytes(JavaThread::satb_mark_queue_offset() +
                                  SATBMarkQueue::byte_offset_of_buf()));

  Label done;
  Label runtime;

  // Is marking still active?
  if (in_bytes(SATBMarkQueue::byte_width_of_active()) == 4) {
    __ cmpl(queue_active, 0);
  } else {
    assert(in_bytes(SATBMarkQueue::byte_width_of_active()) == 1, "Assumption");
    __ cmpb(queue_active, 0);
  }
  __ jcc(Assembler::equal, done);

  // Can we store original value in the thread's buffer?

  __ movptr(tmp, queue_index);
  __ testptr(tmp, tmp);
  __ jcc(Assembler::zero, runtime);
  __ subptr(tmp, wordSize);
  __ movptr(queue_index, tmp);
  __ addptr(tmp, buffer);

  // prev_val (rax)
  __ load_parameter(0, pre_val);
  __ movptr(Address(tmp, 0), pre_val);
  __ jmp(done);

  __ bind(runtime);

  __ save_live_registers_no_oop_map(3, true);

  // load the pre-value
  __ load_parameter(0, rcx);
  __ call_VM_leaf(CAST_FROM_FN_PTR(address, G1BarrierSet::g1_wb_pre), rcx, thread);

  __ restore_live_registers(true);

  __ bind(done);

  __ pop(rdx);
  __ pop(rax);

  __ epilogue();
}

void G1BSCodeGen::generate_c1_post_barrier_runtime_stub(StubAssembler* sasm) {
  __ prologue("g1_post_barrier", false);

  // arg0: store_address
  Address store_addr(rbp, 2*BytesPerWord);

  CardTableModRefBS* ct =
    barrier_set_cast<CardTableModRefBS>(Universe::heap()->barrier_set());
  assert(sizeof(*ct->card_table()->byte_map_base()) == sizeof(jbyte), "adjust this code");

  Label done;
  Label enqueued;
  Label runtime;

  // At this point we know new_value is non-NULL and the new_value crosses regions.
  // Must check to see if card is already dirty

  const Register thread = NOT_LP64(rax) LP64_ONLY(r15_thread);

  Address queue_index(thread, in_bytes(JavaThread::dirty_card_queue_offset() +
                                       DirtyCardQueue::byte_offset_of_index()));
  Address buffer(thread, in_bytes(JavaThread::dirty_card_queue_offset() +
                                  DirtyCardQueue::byte_offset_of_buf()));

  __ push(rax);
  __ push(rcx);

  const Register cardtable = rax;
  const Register card_addr = rcx;

  __ load_parameter(0, card_addr);
  __ shrptr(card_addr, CardTable::card_shift);
  // Do not use ExternalAddress to load 'byte_map_base', since 'byte_map_base' is NOT
  // a valid address and therefore is not properly handled by the relocation code.
  __ movptr(cardtable, (intptr_t)ct->card_table()->byte_map_base());
  __ addptr(card_addr, cardtable);

  NOT_LP64(__ get_thread(thread);)

  __ cmpb(Address(card_addr, 0), (int)G1CardTable::g1_young_card_val());
  __ jcc(Assembler::equal, done);

  __ membar(Assembler::Membar_mask_bits(Assembler::StoreLoad));
  __ cmpb(Address(card_addr, 0), (int)G1CardTable::dirty_card_val());
  __ jcc(Assembler::equal, done);

  // storing region crossing non-NULL, card is clean.
  // dirty card and log.

  __ movb(Address(card_addr, 0), (int)G1CardTable::dirty_card_val());

  const Register tmp = rdx;
  __ push(rdx);

  __ movptr(tmp, queue_index);
  __ testptr(tmp, tmp);
  __ jcc(Assembler::zero, runtime);
  __ subptr(tmp, wordSize);
  __ movptr(queue_index, tmp);
  __ addptr(tmp, buffer);
  __ movptr(Address(tmp, 0), card_addr);
  __ jmp(enqueued);

  __ bind(runtime);

  __ save_live_registers_no_oop_map(3, true);

  __ call_VM_leaf(CAST_FROM_FN_PTR(address, G1BarrierSet::g1_wb_post), card_addr, thread);

  __ restore_live_registers(true);

  __ bind(enqueued);
  __ pop(rdx);

  __ bind(done);
  __ pop(rcx);
  __ pop(rax);

  __ epilogue();
}

#undef __