view src/hotspot/share/code/compiledIC.hpp @ 51020:ef980b9ac191

8203837: Split nmethod unloading from inline cache cleaning Summary: Refactor cleaning inline caches to after GC do_unloading. Reviewed-by: thartmann, eosterlund
author coleenp
date Wed, 02 May 2018 11:28:49 -0400
parents 1dab70e20292
children f6641fcf7b7e
line wrap: on
line source
/*
 * Copyright (c) 1997, 2017, 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.
 *
 */

#ifndef SHARE_VM_CODE_COMPILEDIC_HPP
#define SHARE_VM_CODE_COMPILEDIC_HPP

#include "code/nativeInst.hpp"
#include "interpreter/linkResolver.hpp"
#include "oops/compiledICHolder.hpp"

//-----------------------------------------------------------------------------
// The CompiledIC represents a compiled inline cache.
//
// In order to make patching of the inline cache MT-safe, we only allow the following
// transitions (when not at a safepoint):
//
//
//         [1] --<--  Clean -->---  [1]
//            /       (null)      \
//           /                     \      /-<-\
//          /          [2]          \    /     \
//      Interpreted  ---------> Monomorphic     | [3]
//  (CompiledICHolder*)            (Klass*)     |
//          \                        /   \     /
//       [4] \                      / [4] \->-/
//            \->-  Megamorphic -<-/
//              (CompiledICHolder*)
//
// The text in parentheses () refers to the value of the inline cache receiver (mov instruction)
//
// The numbers in square brackets refer to the kind of transition:
// [1]: Initial fixup. Receiver it found from debug information
// [2]: Compilation of a method
// [3]: Recompilation of a method (note: only entry is changed. The Klass* must stay the same)
// [4]: Inline cache miss. We go directly to megamorphic call.
//
// The class automatically inserts transition stubs (using the InlineCacheBuffer) when an MT-unsafe
// transition is made to a stub.
//
class CompiledIC;
class ICStub;

class CompiledICInfo : public StackObj {
 private:
  address _entry;              // entry point for call
  void*   _cached_value;         // Value of cached_value (either in stub or inline cache)
  bool    _is_icholder;          // Is the cached value a CompiledICHolder*
  bool    _is_optimized;       // it is an optimized virtual call (i.e., can be statically bound)
  bool    _to_interpreter;     // Call it to interpreter
  bool    _to_aot;             // Call it to aot code
  bool    _release_icholder;
 public:
  address entry() const        { return _entry; }
  Metadata*    cached_metadata() const         { assert(!_is_icholder, ""); return (Metadata*)_cached_value; }
  CompiledICHolder*    claim_cached_icholder() {
    assert(_is_icholder, "");
    assert(_cached_value != NULL, "must be non-NULL");
    _release_icholder = false;
    CompiledICHolder* icholder = (CompiledICHolder*)_cached_value;
    icholder->claim();
    return icholder;
  }
  bool    is_optimized() const { return _is_optimized; }
  bool  to_interpreter() const { return _to_interpreter; }
  bool          to_aot() const { return _to_aot; }

  void set_compiled_entry(address entry, Klass* klass, bool is_optimized) {
    _entry      = entry;
    _cached_value = (void*)klass;
    _to_interpreter = false;
    _to_aot = false;
    _is_icholder = false;
    _is_optimized = is_optimized;
    _release_icholder = false;
  }

  void set_interpreter_entry(address entry, Method* method) {
    _entry      = entry;
    _cached_value = (void*)method;
    _to_interpreter = true;
    _to_aot = false;
    _is_icholder = false;
    _is_optimized = true;
    _release_icholder = false;
  }

  void set_aot_entry(address entry, Method* method) {
    _entry      = entry;
    _cached_value = (void*)method;
    _to_interpreter = false;
    _to_aot = true;
    _is_icholder = false;
    _is_optimized = true;
    _release_icholder = false;
  }

  void set_icholder_entry(address entry, CompiledICHolder* icholder) {
    _entry      = entry;
    _cached_value = (void*)icholder;
    _to_interpreter = true;
    _to_aot = false;
    _is_icholder = true;
    _is_optimized = false;
    _release_icholder = true;
  }

  CompiledICInfo(): _entry(NULL), _cached_value(NULL), _is_icholder(false),
                    _to_interpreter(false), _to_aot(false), _is_optimized(false), _release_icholder(false) {
  }
  ~CompiledICInfo() {
    // In rare cases the info is computed but not used, so release any
    // CompiledICHolder* that was created
    if (_release_icholder) {
      assert(_is_icholder, "must be");
      CompiledICHolder* icholder = (CompiledICHolder*)_cached_value;
      icholder->claim();
      delete icholder;
    }
  }
};

class NativeCallWrapper: public ResourceObj {
public:
  virtual address destination() const = 0;
  virtual address instruction_address() const = 0;
  virtual address next_instruction_address() const = 0;
  virtual address return_address() const = 0;
  virtual address get_resolve_call_stub(bool is_optimized) const = 0;
  virtual void set_destination_mt_safe(address dest) = 0;
  virtual void set_to_interpreted(const methodHandle& method, CompiledICInfo& info) = 0;
  virtual void verify() const = 0;
  virtual void verify_resolve_call(address dest) const = 0;

  virtual bool is_call_to_interpreted(address dest) const = 0;
  virtual bool is_safe_for_patching() const = 0;

  virtual NativeInstruction* get_load_instruction(virtual_call_Relocation* r) const = 0;

  virtual void *get_data(NativeInstruction* instruction) const = 0;
  virtual void set_data(NativeInstruction* instruction, intptr_t data) = 0;
};

class CompiledIC: public ResourceObj {
  friend class InlineCacheBuffer;
  friend class ICStub;

 private:
  NativeCallWrapper* _call;
  NativeInstruction* _value;    // patchable value cell for this IC
  bool          _is_optimized;  // an optimized virtual call (i.e., no compiled IC)
  CompiledMethod* _method;

  CompiledIC(CompiledMethod* cm, NativeCall* ic_call);
  CompiledIC(RelocIterator* iter);

  void initialize_from_iter(RelocIterator* iter);

  static bool is_icholder_entry(address entry);

  // low-level inline-cache manipulation. Cannot be accessed directly, since it might not be MT-safe
  // to change an inline-cache. These changes the underlying inline-cache directly. They *newer* make
  // changes to a transition stub.
  void internal_set_ic_destination(address entry_point, bool is_icstub, void* cache, bool is_icholder);
  void set_ic_destination(ICStub* stub);
  void set_ic_destination(address entry_point) {
    assert(_is_optimized, "use set_ic_destination_and_value instead");
    internal_set_ic_destination(entry_point, false, NULL, false);
  }
  // This only for use by ICStubs where the type of the value isn't known
  void set_ic_destination_and_value(address entry_point, void* value) {
    internal_set_ic_destination(entry_point, false, value, is_icholder_entry(entry_point));
  }
  void set_ic_destination_and_value(address entry_point, Metadata* value) {
    internal_set_ic_destination(entry_point, false, value, false);
  }
  void set_ic_destination_and_value(address entry_point, CompiledICHolder* value) {
    internal_set_ic_destination(entry_point, false, value, true);
  }

  // Reads the location of the transition stub. This will fail with an assertion, if no transition stub is
  // associated with the inline cache.
  address stub_address() const;
  bool is_in_transition_state() const;  // Use InlineCacheBuffer

 public:
  // conversion (machine PC to CompiledIC*)
  friend CompiledIC* CompiledIC_before(CompiledMethod* nm, address return_addr);
  friend CompiledIC* CompiledIC_at(CompiledMethod* nm, address call_site);
  friend CompiledIC* CompiledIC_at(Relocation* call_site);
  friend CompiledIC* CompiledIC_at(RelocIterator* reloc_iter);

  // This is used to release CompiledICHolder*s from nmethods that
  // are about to be freed.  The callsite might contain other stale
  // values of other kinds so it must be careful.
  static void cleanup_call_site(virtual_call_Relocation* call_site, const CompiledMethod* cm);
  static bool is_icholder_call_site(virtual_call_Relocation* call_site, const CompiledMethod* cm);

  // Return the cached_metadata/destination associated with this inline cache. If the cache currently points
  // to a transition stub, it will read the values from the transition stub.
  void* cached_value() const;
  CompiledICHolder* cached_icholder() const {
    assert(is_icholder_call(), "must be");
    return (CompiledICHolder*) cached_value();
  }
  Metadata* cached_metadata() const {
    assert(!is_icholder_call(), "must be");
    return (Metadata*) cached_value();
  }

  void* get_data() const {
    return _call->get_data(_value);
  }

  void set_data(intptr_t data) {
    _call->set_data(_value, data);
  }

  address ic_destination() const;

  bool is_optimized() const   { return _is_optimized; }

  // State
  bool is_clean() const;
  bool is_megamorphic() const;
  bool is_call_to_compiled() const;
  bool is_call_to_interpreted() const;

  bool is_icholder_call() const;

  address end_of_call() { return  _call->return_address(); }

  // MT-safe patching of inline caches. Note: Only safe to call is_xxx when holding the CompiledIC_ock
  // so you are guaranteed that no patching takes place. The same goes for verify.
  //
  // Note: We do not provide any direct access to the stub code, to prevent parts of the code
  // to manipulate the inline cache in MT-unsafe ways.
  //
  // They all takes a TRAP argument, since they can cause a GC if the inline-cache buffer is full.
  //
  void set_to_clean(bool in_use = true);
  void set_to_monomorphic(CompiledICInfo& info);
  void clear_ic_stub();

  // Returns true if successful and false otherwise. The call can fail if memory
  // allocation in the code cache fails.
  bool set_to_megamorphic(CallInfo* call_info, Bytecodes::Code bytecode, TRAPS);

  static void compute_monomorphic_entry(const methodHandle& method, Klass* receiver_klass,
                                        bool is_optimized, bool static_bound, bool caller_is_nmethod,
                                        CompiledICInfo& info, TRAPS);

  // Location
  address instruction_address() const { return _call->instruction_address(); }

  // Misc
  void print()             PRODUCT_RETURN;
  void print_compiled_ic() PRODUCT_RETURN;
  void verify()            PRODUCT_RETURN;
};

inline CompiledIC* CompiledIC_before(CompiledMethod* nm, address return_addr) {
  CompiledIC* c_ic = new CompiledIC(nm, nativeCall_before(return_addr));
  c_ic->verify();
  return c_ic;
}

inline CompiledIC* CompiledIC_at(CompiledMethod* nm, address call_site) {
  CompiledIC* c_ic = new CompiledIC(nm, nativeCall_at(call_site));
  c_ic->verify();
  return c_ic;
}

inline CompiledIC* CompiledIC_at(Relocation* call_site) {
  assert(call_site->type() == relocInfo::virtual_call_type ||
         call_site->type() == relocInfo::opt_virtual_call_type, "wrong reloc. info");
  CompiledIC* c_ic = new CompiledIC(call_site->code(), nativeCall_at(call_site->addr()));
  c_ic->verify();
  return c_ic;
}

inline CompiledIC* CompiledIC_at(RelocIterator* reloc_iter) {
  assert(reloc_iter->type() == relocInfo::virtual_call_type ||
      reloc_iter->type() == relocInfo::opt_virtual_call_type, "wrong reloc. info");
  CompiledIC* c_ic = new CompiledIC(reloc_iter);
  c_ic->verify();
  return c_ic;
}

//-----------------------------------------------------------------------------
// The CompiledStaticCall represents a call to a static method in the compiled
//
// Transition diagram of a static call site is somewhat simpler than for an inlined cache:
//
//
//           -----<----- Clean ----->-----
//          /                             \
//         /                               \
//    compilled code <------------> interpreted code
//
//  Clean:            Calls directly to runtime method for fixup
//  Compiled code:    Calls directly to compiled code
//  Interpreted code: Calls to stub that set Method* reference
//
//

class StaticCallInfo {
 private:
  address      _entry;          // Entrypoint
  methodHandle _callee;         // Callee (used when calling interpreter)
  bool         _to_interpreter; // call to interpreted method (otherwise compiled)
  bool         _to_aot;         // call to aot method (otherwise compiled)

  friend class CompiledStaticCall;
  friend class CompiledDirectStaticCall;
  friend class CompiledPltStaticCall;
 public:
  address      entry() const    { return _entry;  }
  methodHandle callee() const   { return _callee; }
};

class CompiledStaticCall : public ResourceObj {
 public:
  // Code
  static address emit_to_interp_stub(CodeBuffer &cbuf, address mark = NULL);
  static int to_interp_stub_size();
  static int to_trampoline_stub_size();
  static int reloc_to_interp_stub();
  static void emit_to_aot_stub(CodeBuffer &cbuf, address mark = NULL);
  static int to_aot_stub_size();
  static int reloc_to_aot_stub();

  // Compute entry point given a method
  static void compute_entry(const methodHandle& m, bool caller_is_nmethod, StaticCallInfo& info);

public:
  // Clean static call (will force resolving on next use)
  virtual address destination() const = 0;

  // Clean static call (will force resolving on next use)
  void set_to_clean(bool in_use = true);

  // Set state. The entry must be the same, as computed by compute_entry.
  // Computation and setting is split up, since the actions are separate during
  // a OptoRuntime::resolve_xxx.
  void set(const StaticCallInfo& info);

  // State
  bool is_clean() const;
  bool is_call_to_compiled() const;
  virtual bool is_call_to_interpreted() const = 0;

  virtual address instruction_address() const = 0;
protected:
  virtual address resolve_call_stub() const = 0;
  virtual void set_destination_mt_safe(address dest) = 0;
#if INCLUDE_AOT
  virtual void set_to_far(const methodHandle& callee, address entry) = 0;
#endif
  virtual void set_to_interpreted(const methodHandle& callee, address entry) = 0;
  virtual const char* name() const = 0;

  void set_to_compiled(address entry);
};

class CompiledDirectStaticCall : public CompiledStaticCall {
private:
  friend class CompiledIC;
  friend class DirectNativeCallWrapper;

  // Also used by CompiledIC
  void set_to_interpreted(const methodHandle& callee, address entry);
#if INCLUDE_AOT
  void set_to_far(const methodHandle& callee, address entry);
#endif
  address instruction_address() const { return _call->instruction_address(); }
  void set_destination_mt_safe(address dest) { _call->set_destination_mt_safe(dest); }

  NativeCall* _call;

  CompiledDirectStaticCall(NativeCall* call) : _call(call) {}

 public:
  static inline CompiledDirectStaticCall* before(address return_addr) {
    CompiledDirectStaticCall* st = new CompiledDirectStaticCall(nativeCall_before(return_addr));
    st->verify();
    return st;
  }

  static inline CompiledDirectStaticCall* at(address native_call) {
    CompiledDirectStaticCall* st = new CompiledDirectStaticCall(nativeCall_at(native_call));
    st->verify();
    return st;
  }

  static inline CompiledDirectStaticCall* at(Relocation* call_site) {
    return at(call_site->addr());
  }

  // Delegation
  address destination() const { return _call->destination(); }

  // State
  virtual bool is_call_to_interpreted() const;
  bool is_call_to_far() const;

  // Stub support
  static address find_stub_for(address instruction, bool is_aot);
  address find_stub(bool is_aot);
  static void set_stub_to_clean(static_stub_Relocation* static_stub);

  // Misc.
  void print()  PRODUCT_RETURN;
  void verify() PRODUCT_RETURN;

 protected:
  virtual address resolve_call_stub() const;
  virtual const char* name() const { return "CompiledDirectStaticCall"; }
};

#endif // SHARE_VM_CODE_COMPILEDIC_HPP