view src/os_cpu/windows_x86/vm/windows_x86_32.ad @ 1886:72a161e62cc4

6991377: G1: race between concurrent refinement and humongous object allocation Summary: There is a race between the concurrent refinement threads and the humongous object allocation that can cause the concurrent refinement threads to corrupt the part of the BOT that it is being initialized by the humongous object allocation operation. The solution is to do the humongous object allocation in careful steps to ensure that the concurrent refinement threads always have a consistent view over the BOT, region contents, and top. The fix includes some very minor tidying up in sparsePRT. Reviewed-by: jcoomes, johnc, ysr
author tonyp
date Sat, 16 Oct 2010 17:12:19 -0400
parents 3e8fbc61cee8
children 9b8ce46870df
line wrap: on
line source

//
// Copyright (c) 1999, 2010, 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.
//
//

// X86 Win32 Architecture Description File

//----------OS-DEPENDENT ENCODING BLOCK-----------------------------------------------------
// This block specifies the encoding classes used by the compiler to output
// byte streams.  Encoding classes generate functions which are called by
// Machine Instruction Nodes in order to generate the bit encoding of the
// instruction.  Operands specify their base encoding interface with the
// interface keyword.  There are currently supported four interfaces,
// REG_INTER, CONST_INTER, MEMORY_INTER, & COND_INTER.  REG_INTER causes an
// operand to generate a function which returns its register number when
// queried.   CONST_INTER causes an operand to generate a function which
// returns the value of the constant when queried.  MEMORY_INTER causes an
// operand to generate four functions which return the Base Register, the
// Index Register, the Scale Value, and the Offset Value of the operand when
// queried.  COND_INTER causes an operand to generate six functions which
// return the encoding code (ie - encoding bits for the instruction)
// associated with each basic boolean condition for a conditional instruction.
// Instructions specify two basic values for encoding.  They use the
// ins_encode keyword to specify their encoding class (which must be one of
// the class names specified in the encoding block), and they use the
// opcode keyword to specify, in order, their primary, secondary, and
// tertiary opcode.  Only the opcode sections which a particular instruction
// needs for encoding need to be specified.
encode %{
  // Build emit functions for each basic byte or larger field in the intel
  // encoding scheme (opcode, rm, sib, immediate), and call them from C++
  // code in the enc_class source block.  Emit functions will live in the
  // main source block for now.  In future, we can generalize this by
  // adding a syntax that specifies the sizes of fields in an order,
  // so that the adlc can build the emit functions automagically

  enc_class tlsencode (eRegP dst, eRegP src) %{
    emit_rm(cbuf, 0x2, $dst$$reg, $src$$reg);
    emit_d32(cbuf, ThreadLocalStorage::get_thread_ptr_offset() );
  %}

  enc_class call_epilog %{
    if( VerifyStackAtCalls ) {
      // Check that stack depth is unchanged: find majik cookie on stack
      int framesize = ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP,-3*VMRegImpl::slots_per_word));
      if(framesize >= 128) {
        emit_opcode(cbuf, 0x81); // cmp [esp+0],0xbadb1ood
        emit_d8(cbuf,0xBC);
        emit_d8(cbuf,0x24);
        emit_d32(cbuf,framesize); // Find majik cookie from ESP
        emit_d32(cbuf, 0xbadb100d);
      }
      else {
        emit_opcode(cbuf, 0x81); // cmp [esp+0],0xbadb1ood
        emit_d8(cbuf,0x7C);
        emit_d8(cbuf,0x24);
        emit_d8(cbuf,framesize); // Find majik cookie from ESP
        emit_d32(cbuf, 0xbadb100d);
      }
      // jmp EQ around INT3
      emit_opcode(cbuf,0x74);
      emit_d8(cbuf,1);
      // Die if stack mismatch
      emit_opcode(cbuf,0xCC);
    }
  %}

%}

// INSTRUCTIONS -- Platform dependent


//----------OS and Locking Instructions----------------------------------------

// The prefix of this name is KNOWN by the ADLC and cannot be changed.
instruct tlsLoadP_prefixLoadP(eRegP t1) %{
  effect(DEF t1);

  format %{ "MOV    $t1,FS:[0x00] "%}
  opcode(0x8B, 0x64);
  ins_encode(OpcS, OpcP, conmemref(t1));
  ins_pipe( ialu_reg_fat );
%}

// This name is KNOWN by the ADLC and cannot be changed.
// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
// for this guy.
// %%% Should do this with a clause like:  bottom_type(TypeRawPtr::BOTTOM);
instruct tlsLoadP(eRegP dst, eRegP t1) %{
  effect(DEF dst, USE t1);

  format %{ "MOV    $dst,[$t1 + TLS::thread_ptr_offset()]" %}
  opcode(0x8B);
  ins_encode(OpcP, tlsencode(dst, t1));
  ins_pipe( ialu_reg_reg_fat );
%}

instruct TLS(eRegP dst) %{
  match(Set dst (ThreadLocal));
  expand %{
    eRegP t1;
    tlsLoadP_prefixLoadP(t1);
    tlsLoadP(dst, t1);
  %}
%}

// Die now
instruct ShouldNotReachHere( )
%{
  match(Halt);
  // Use the following format syntax
  format %{ "INT3   ; ShouldNotReachHere" %}
  opcode(0xCC);
  ins_encode(OpcP);
  ins_pipe( pipe_slow );
%}

//
// Platform dependent source
//
source %{

// emit an interrupt that is caught by the debugger
void emit_break(CodeBuffer &cbuf) {
  cbuf.insts()->emit_int8((unsigned char) 0xcc);
}

void MachBreakpointNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
  emit_break(cbuf);
}


uint MachBreakpointNode::size(PhaseRegAlloc *ra_) const {
  return 1;
}


%}