Quelle codeCache.cpp Sprache: C

/*
* Copyright (c) 1997, 2022, 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 "code/codeBlob.hpp"
#include "code/codeCache.hpp"
#include "code/codeHeapState.hpp"
#include "code/compiledIC.hpp"
#include "code/dependencies.hpp"
#include "code/dependencyContext.hpp"
#include "code/icBuffer.hpp"
#include "code/nmethod.hpp"
#include "code/pcDesc.hpp"
#include "compiler/compilationPolicy.hpp"
#include "compiler/compileBroker.hpp"
#include "compiler/compilerDefinitions.inline.hpp"
#include "compiler/oopMap.hpp"
#include "gc/shared/barrierSetNMethod.hpp"
#include "gc/shared/collectedHeap.hpp"
#include "jfr/jfrEvents.hpp"
#include "jvm_io.h"
#include "logging/log.hpp"
#include "logging/logStream.hpp"
#include "memory/allocation.inline.hpp"
#include "memory/iterator.hpp"
#include "memory/resourceArea.hpp"
#include "memory/universe.hpp"
#include "oops/method.inline.hpp"
#include "oops/objArrayOop.hpp"
#include "oops/oop.inline.hpp"
#include "oops/verifyOopClosure.hpp"
#include "runtime/arguments.hpp"
#include "runtime/atomic.hpp"
#include "runtime/deoptimization.hpp"
#include "runtime/globals_extension.hpp"
#include "runtime/handles.inline.hpp"
#include "runtime/icache.hpp"
#include "runtime/init.hpp"
#include "runtime/java.hpp"
#include "runtime/mutexLocker.hpp"
#include "runtime/os.inline.hpp"
#include "runtime/safepointVerifiers.hpp"
#include "runtime/vmThread.hpp"
#include "services/memoryService.hpp"
#include "utilities/align.hpp"
#include "utilities/vmError.hpp"
#include "utilities/xmlstream.hpp"
#ifdef COMPILER1
#include "c1/c1_Compilation.hpp"
#include "c1/c1_Compiler.hpp"
#endif
#ifdef COMPILER2
#include "opto/c2compiler.hpp"
#include "opto/compile.hpp"
#include "opto/node.hpp"
#endif

// Helper class for printing in CodeCache
class CodeBlob_sizes {
private:
  int count;
  int total_size;
  int header_size;
  int code_size;
  int stub_size;
  int relocation_size;
  int scopes_oop_size;
  int scopes_metadata_size;
  int scopes_data_size;
  int scopes_pcs_size;

public:
  CodeBlob_sizes() {
    count            = 0;
    total_size       = 0;
    header_size      = 0;
    code_size        = 0;
    stub_size        = 0;
    relocation_size  = 0;
    scopes_oop_size  = 0;
    scopes_metadata_size  = 0;
    scopes_data_size = 0;
    scopes_pcs_size  = 0;
  }

  int total() const                              { return total_size; }
  bool is_empty() const                          { return count == 0; }

  void print(const char* title) const {
    if (is_empty()) {
      tty->print_cr(" #%d %s = %dK",
                    count,
                    title,
                    total()                 / (int)K);
    } else {
      tty->print_cr(" #%d %s = %dK (hdr %dK %d%%, loc %dK %d%%, code %dK %d%%, stub %dK %d%%, [oops %dK %d%%, metadata %dK %d%%, data %dK %d%%, pcs %dK %d%%])",
                    count,
                    title,
                    total()                 / (int)K,
                    header_size             / (int)K,
                    header_size             * 100 / total_size,
                    relocation_size         / (int)K,
                    relocation_size         * 100 / total_size,
                    code_size               / (int)K,
                    code_size               * 100 / total_size,
                    stub_size               / (int)K,
                    stub_size               * 100 / total_size,
                    scopes_oop_size         / (int)K,
                    scopes_oop_size         * 100 / total_size,
                    scopes_metadata_size    / (int)K,
                    scopes_metadata_size    * 100 / total_size,
                    scopes_data_size        / (int)K,
                    scopes_data_size        * 100 / total_size,
                    scopes_pcs_size         / (int)K,
                    scopes_pcs_size         * 100 / total_size);
    }
  }

  void add(CodeBlob* cb) {
    count++;
    total_size       += cb->size();
    header_size      += cb->header_size();
    relocation_size  += cb->relocation_size();
    if (cb->is_nmethod()) {
      nmethod* nm = cb->as_nmethod_or_null();
      code_size        += nm->insts_size();
      stub_size        += nm->stub_size();

      scopes_oop_size  += nm->oops_size();
      scopes_metadata_size  += nm->metadata_size();
      scopes_data_size += nm->scopes_data_size();
      scopes_pcs_size  += nm->scopes_pcs_size();
    } else {
      code_size        += cb->code_size();
    }
  }
};

// Iterate over all CodeHeaps
#define FOR_ALL_HEAPS(heap) for (GrowableArrayIterator<CodeHeap*> heap = _heaps->begin(); heap != _heaps->end(); ++heap)
#define FOR_ALL_NMETHOD_HEAPS(heap) for (GrowableArrayIterator<CodeHeap*> heap = _nmethod_heaps->begin(); heap != _nmethod_heaps->end(); ++heap)
#define FOR_ALL_ALLOCABLE_HEAPS(heap) for (GrowableArrayIterator<CodeHeap*> heap = _allocable_heaps->begin(); heap != _allocable_heaps->end(); ++heap)

// Iterate over all CodeBlobs (cb) on the given CodeHeap
#define FOR_ALL_BLOBS(cb, heap) for (CodeBlob* cb = first_blob(heap); cb != NULL; cb = next_blob(heap, cb))

address CodeCache::_low_bound = 0;
address CodeCache::_high_bound = 0;
int CodeCache::_number_of_nmethods_with_dependencies = 0;
ExceptionCache* volatile CodeCache::_exception_cache_purge_list = NULL;

// Initialize arrays of CodeHeap subsets
GrowableArray<CodeHeap*>* CodeCache::_heaps = new(mtCode) GrowableArray<CodeHeap*> (static_cast<int>(CodeBlobType::All), mtCode);
GrowableArray<CodeHeap*>* CodeCache::_compiled_heaps = new(mtCode) GrowableArray<CodeHeap*> (static_cast<int>(CodeBlobType::All), mtCode);
GrowableArray<CodeHeap*>* CodeCache::_nmethod_heaps = new(mtCode) GrowableArray<CodeHeap*> (static_cast<int>(CodeBlobType::All), mtCode);
GrowableArray<CodeHeap*>* CodeCache::_allocable_heaps = new(mtCode) GrowableArray<CodeHeap*> (static_cast<int>(CodeBlobType::All), mtCode);

void CodeCache::check_heap_sizes(size_t non_nmethod_size, size_t profiled_size, size_t non_profiled_size, size_t cache_size, bool all_set) {
  size_t total_size = non_nmethod_size + profiled_size + non_profiled_size;
  // Prepare error message
  const char* error = "Invalid code heap sizes";
  err_msg message("NonNMethodCodeHeapSize (" SIZE_FORMAT "K) + ProfiledCodeHeapSize (" SIZE_FORMAT "K)"
                  " + NonProfiledCodeHeapSize (" SIZE_FORMAT "K) = " SIZE_FORMAT "K",
          non_nmethod_size/K, profiled_size/K, non_profiled_size/K, total_size/K);

  if (total_size > cache_size) {
    // Some code heap sizes were explicitly set: total_size must be <= cache_size
    message.append(" is greater than ReservedCodeCacheSize (" SIZE_FORMAT "K).", cache_size/K);
    vm_exit_during_initialization(error, message);
  } else if (all_set && total_size != cache_size) {
    // All code heap sizes were explicitly set: total_size must equal cache_size
    message.append(" is not equal to ReservedCodeCacheSize (" SIZE_FORMAT "K).", cache_size/K);
    vm_exit_during_initialization(error, message);
  }
}

void CodeCache::initialize_heaps() {
  bool non_nmethod_set      = FLAG_IS_CMDLINE(NonNMethodCodeHeapSize);
  bool profiled_set         = FLAG_IS_CMDLINE(ProfiledCodeHeapSize);
  bool non_profiled_set     = FLAG_IS_CMDLINE(NonProfiledCodeHeapSize);
  size_t min_size           = os::vm_page_size();
  size_t cache_size         = ReservedCodeCacheSize;
  size_t non_nmethod_size   = NonNMethodCodeHeapSize;
  size_t profiled_size      = ProfiledCodeHeapSize;
  size_t non_profiled_size  = NonProfiledCodeHeapSize;
  // Check if total size set via command line flags exceeds the reserved size
  check_heap_sizes((non_nmethod_set  ? non_nmethod_size  : min_size),
                   (profiled_set     ? profiled_size     : min_size),
                   (non_profiled_set ? non_profiled_size : min_size),
                   cache_size,
                   non_nmethod_set && profiled_set && non_profiled_set);

  // Determine size of compiler buffers
  size_t code_buffers_size = 0;
#ifdef COMPILER1
  // C1 temporary code buffers (see Compiler::init_buffer_blob())
  const int c1_count = CompilationPolicy::c1_count();
  code_buffers_size += c1_count * Compiler::code_buffer_size();
#endif
#ifdef COMPILER2
  // C2 scratch buffers (see Compile::init_scratch_buffer_blob())
  const int c2_count = CompilationPolicy::c2_count();
  // Initial size of constant table (this may be increased if a compiled method needs more space)
  code_buffers_size += c2_count * C2Compiler::initial_code_buffer_size();
#endif

  // Increase default non_nmethod_size to account for compiler buffers
  if (!non_nmethod_set) {
    non_nmethod_size += code_buffers_size;
  }
  // Calculate default CodeHeap sizes if not set by user
  if (!non_nmethod_set && !profiled_set && !non_profiled_set) {
    // Check if we have enough space for the non-nmethod code heap
    if (cache_size > non_nmethod_size) {
      // Use the default value for non_nmethod_size and one half of the
      // remaining size for non-profiled and one half for profiled methods
      size_t remaining_size = cache_size - non_nmethod_size;
      profiled_size = remaining_size / 2;
      non_profiled_size = remaining_size - profiled_size;
    } else {
      // Use all space for the non-nmethod heap and set other heaps to minimal size
      non_nmethod_size = cache_size - 2 * min_size;
      profiled_size = min_size;
      non_profiled_size = min_size;
    }
  } else if (!non_nmethod_set || !profiled_set || !non_profiled_set) {
    // The user explicitly set some code heap sizes. Increase or decrease the (default)
    // sizes of the other code heaps accordingly. First adapt non-profiled and profiled
    // code heap sizes and then only change non-nmethod code heap size if still necessary.
    intx diff_size = cache_size - (non_nmethod_size + profiled_size + non_profiled_size);
    if (non_profiled_set) {
      if (!profiled_set) {
        // Adapt size of profiled code heap
        if (diff_size < 0 && ((intx)profiled_size + diff_size) <= 0) {
          // Not enough space available, set to minimum size
          diff_size += profiled_size - min_size;
          profiled_size = min_size;
        } else {
          profiled_size += diff_size;
          diff_size = 0;
        }
      }
    } else if (profiled_set) {
      // Adapt size of non-profiled code heap
      if (diff_size < 0 && ((intx)non_profiled_size + diff_size) <= 0) {
        // Not enough space available, set to minimum size
        diff_size += non_profiled_size - min_size;
        non_profiled_size = min_size;
      } else {
        non_profiled_size += diff_size;
        diff_size = 0;
      }
    } else if (non_nmethod_set) {
      // Distribute remaining size between profiled and non-profiled code heaps
      diff_size = cache_size - non_nmethod_size;
      profiled_size = diff_size / 2;
      non_profiled_size = diff_size - profiled_size;
      diff_size = 0;
    }
    if (diff_size != 0) {
      // Use non-nmethod code heap for remaining space requirements
      assert(!non_nmethod_set && ((intx)non_nmethod_size + diff_size) > 0, "sanity");
      non_nmethod_size += diff_size;
    }
  }

  // We do not need the profiled CodeHeap, use all space for the non-profiled CodeHeap
  if (!heap_available(CodeBlobType::MethodProfiled)) {
    non_profiled_size += profiled_size;
    profiled_size = 0;
  }
  // We do not need the non-profiled CodeHeap, use all space for the non-nmethod CodeHeap
  if (!heap_available(CodeBlobType::MethodNonProfiled)) {
    non_nmethod_size += non_profiled_size;
    non_profiled_size = 0;
  }
  // Make sure we have enough space for VM internal code
  uint min_code_cache_size = CodeCacheMinimumUseSpace DEBUG_ONLY(* 3);
  if (non_nmethod_size < min_code_cache_size) {
    vm_exit_during_initialization(err_msg(
        "Not enough space in non-nmethod code heap to run VM: " SIZE_FORMAT "K < " SIZE_FORMAT "K",
        non_nmethod_size/K, min_code_cache_size/K));
  }

  // Verify sizes and update flag values
  assert(non_profiled_size + profiled_size + non_nmethod_size == cache_size, "Invalid code heap sizes");
  FLAG_SET_ERGO(NonNMethodCodeHeapSize, non_nmethod_size);
  FLAG_SET_ERGO(ProfiledCodeHeapSize, profiled_size);
  FLAG_SET_ERGO(NonProfiledCodeHeapSize, non_profiled_size);

  // If large page support is enabled, align code heaps according to large
  // page size to make sure that code cache is covered by large pages.
  const size_t alignment = MAX2(page_size(false, 8), (size_t) os::vm_allocation_granularity());
  non_nmethod_size = align_up(non_nmethod_size, alignment);
  profiled_size    = align_down(profiled_size, alignment);
  non_profiled_size = align_down(non_profiled_size, alignment);

  // Reserve one continuous chunk of memory for CodeHeaps and split it into
  // parts for the individual heaps. The memory layout looks like this:
  // ---------- high -----------
  //    Non-profiled nmethods
  //         Non-nmethods
  //      Profiled nmethods
  // ---------- low ------------
  ReservedCodeSpace rs = reserve_heap_memory(cache_size);
  ReservedSpace profiled_space      = rs.first_part(profiled_size);
  ReservedSpace rest                = rs.last_part(profiled_size);
  ReservedSpace non_method_space    = rest.first_part(non_nmethod_size);
  ReservedSpace non_profiled_space  = rest.last_part(non_nmethod_size);

  // Non-nmethods (stubs, adapters, ...)
  add_heap(non_method_space, "CodeHeap 'non-nmethods'", CodeBlobType::NonNMethod);
  // Tier 2 and tier 3 (profiled) methods
  add_heap(profiled_space, "CodeHeap 'profiled nmethods'", CodeBlobType::MethodProfiled);
  // Tier 1 and tier 4 (non-profiled) methods and native methods
  add_heap(non_profiled_space, "CodeHeap 'non-profiled nmethods'", CodeBlobType::MethodNonProfiled);
}

size_t CodeCache::page_size(bool aligned, size_t min_pages) {
  if (os::can_execute_large_page_memory()) {
    if (InitialCodeCacheSize < ReservedCodeCacheSize) {
      // Make sure that the page size allows for an incremental commit of the reserved space
      min_pages = MAX2(min_pages, (size_t)8);
    }
    return aligned ? os::page_size_for_region_aligned(ReservedCodeCacheSize, min_pages) :
                     os::page_size_for_region_unaligned(ReservedCodeCacheSize, min_pages);
  } else {
    return os::vm_page_size();
  }
}

ReservedCodeSpace CodeCache::reserve_heap_memory(size_t size) {
  // Align and reserve space for code cache
  const size_t rs_ps = page_size();
  const size_t rs_align = MAX2(rs_ps, (size_t) os::vm_allocation_granularity());
  const size_t rs_size = align_up(size, rs_align);
  ReservedCodeSpace rs(rs_size, rs_align, rs_ps);
  if (!rs.is_reserved()) {
    vm_exit_during_initialization(err_msg("Could not reserve enough space for code cache (" SIZE_FORMAT "K)",
                                          rs_size/K));
  }

  // Initialize bounds
  _low_bound = (address)rs.base();
  _high_bound = _low_bound + rs.size();
  return rs;
}

// Heaps available for allocation
bool CodeCache::heap_available(CodeBlobType code_blob_type) {
  if (!SegmentedCodeCache) {
    // No segmentation: use a single code heap
    return (code_blob_type == CodeBlobType::All);
  } else if (CompilerConfig::is_interpreter_only()) {
    // Interpreter only: we don't need any method code heaps
    return (code_blob_type == CodeBlobType::NonNMethod);
  } else if (CompilerConfig::is_c1_profiling()) {
    // Tiered compilation: use all code heaps
    return (code_blob_type < CodeBlobType::All);
  } else {
    // No TieredCompilation: we only need the non-nmethod and non-profiled code heap
    return (code_blob_type == CodeBlobType::NonNMethod) ||
           (code_blob_type == CodeBlobType::MethodNonProfiled);
  }
}

const char* CodeCache::get_code_heap_flag_name(CodeBlobType code_blob_type) {
  switch(code_blob_type) {
  case CodeBlobType::NonNMethod:
    return "NonNMethodCodeHeapSize";
    break;
  case CodeBlobType::MethodNonProfiled:
    return "NonProfiledCodeHeapSize";
    break;
  case CodeBlobType::MethodProfiled:
    return "ProfiledCodeHeapSize";
    break;
  default:
    ShouldNotReachHere();
    return NULL;
  }
}

int CodeCache::code_heap_compare(CodeHeap* const &lhs, CodeHeap* const &rhs) {
  if (lhs->code_blob_type() == rhs->code_blob_type()) {
    return (lhs > rhs) ? 1 : ((lhs < rhs) ? -1 : 0);
  } else {
    return static_cast<int>(lhs->code_blob_type()) - static_cast<int>(rhs->code_blob_type());
  }
}

void CodeCache::add_heap(CodeHeap* heap) {
  assert(!Universe::is_fully_initialized(), "late heap addition?");

  _heaps->insert_sorted<code_heap_compare>(heap);

  CodeBlobType type = heap->code_blob_type();
  if (code_blob_type_accepts_compiled(type)) {
    _compiled_heaps->insert_sorted<code_heap_compare>(heap);
  }
  if (code_blob_type_accepts_nmethod(type)) {
    _nmethod_heaps->insert_sorted<code_heap_compare>(heap);
  }
  if (code_blob_type_accepts_allocable(type)) {
    _allocable_heaps->insert_sorted<code_heap_compare>(heap);
  }
}

void CodeCache::add_heap(ReservedSpace rs, const char* name, CodeBlobType code_blob_type) {
  // Check if heap is needed
  if (!heap_available(code_blob_type)) {
    return;
  }

  // Create CodeHeap
  CodeHeap* heap = new CodeHeap(name, code_blob_type);
  add_heap(heap);

  // Reserve Space
  size_t size_initial = MIN2((size_t)InitialCodeCacheSize, rs.size());
  size_initial = align_up(size_initial, os::vm_page_size());
  if (!heap->reserve(rs, size_initial, CodeCacheSegmentSize)) {
    vm_exit_during_initialization(err_msg("Could not reserve enough space in %s (" SIZE_FORMAT "K)",
                                          heap->name(), size_initial/K));
  }

  // Register the CodeHeap
  MemoryService::add_code_heap_memory_pool(heap, name);
}

CodeHeap* CodeCache::get_code_heap_containing(void* start) {
  FOR_ALL_HEAPS(heap) {
    if ((*heap)->contains(start)) {
      return *heap;
    }
  }
  return NULL;
}

CodeHeap* CodeCache::get_code_heap(const CodeBlob* cb) {
  assert(cb != NULL, "CodeBlob is null");
  FOR_ALL_HEAPS(heap) {
    if ((*heap)->contains_blob(cb)) {
      return *heap;
    }
  }
  ShouldNotReachHere();
  return NULL;
}

CodeHeap* CodeCache::get_code_heap(CodeBlobType code_blob_type) {
  FOR_ALL_HEAPS(heap) {
    if ((*heap)->accepts(code_blob_type)) {
      return *heap;
    }
  }
  return NULL;
}

CodeBlob* CodeCache::first_blob(CodeHeap* heap) {
  assert_locked_or_safepoint(CodeCache_lock);
  assert(heap != NULL, "heap is null");
  return (CodeBlob*)heap->first();
}

CodeBlob* CodeCache::first_blob(CodeBlobType code_blob_type) {
  if (heap_available(code_blob_type)) {
    return first_blob(get_code_heap(code_blob_type));
  } else {
    return NULL;
  }
}

CodeBlob* CodeCache::next_blob(CodeHeap* heap, CodeBlob* cb) {
  assert_locked_or_safepoint(CodeCache_lock);
  assert(heap != NULL, "heap is null");
  return (CodeBlob*)heap->next(cb);
}

/**
* Do not seize the CodeCache lock here--if the caller has not
* already done so, we are going to lose bigtime, since the code
* cache will contain a garbage CodeBlob until the caller can
* run the constructor for the CodeBlob subclass he is busy
* instantiating.
*/
CodeBlob* CodeCache::allocate(int size, CodeBlobType code_blob_type, bool handle_alloc_failure, CodeBlobType orig_code_blob_type) {
  assert_locked_or_safepoint(CodeCache_lock);
  assert(size > 0, "Code cache allocation request must be > 0 but is %d", size);
  if (size <= 0) {
    return NULL;
  }
  CodeBlob* cb = NULL;

  // Get CodeHeap for the given CodeBlobType
  CodeHeap* heap = get_code_heap(code_blob_type);
  assert(heap != NULL, "heap is null");

  while (true) {
    cb = (CodeBlob*)heap->allocate(size);
    if (cb != NULL) break;
    if (!heap->expand_by(CodeCacheExpansionSize)) {
      // Save original type for error reporting
      if (orig_code_blob_type == CodeBlobType::All) {
        orig_code_blob_type = code_blob_type;
      }
      // Expansion failed
      if (SegmentedCodeCache) {
        // Fallback solution: Try to store code in another code heap.
        // NonNMethod -> MethodNonProfiled -> MethodProfiled (-> MethodNonProfiled)
        CodeBlobType type = code_blob_type;
        switch (type) {
        case CodeBlobType::NonNMethod:
          type = CodeBlobType::MethodNonProfiled;
          break;
        case CodeBlobType::MethodNonProfiled:
          type = CodeBlobType::MethodProfiled;
          break;
        case CodeBlobType::MethodProfiled:
          // Avoid loop if we already tried that code heap
          if (type == orig_code_blob_type) {
            type = CodeBlobType::MethodNonProfiled;
          }
          break;
        default:
          break;
        }
        if (type != code_blob_type && type != orig_code_blob_type && heap_available(type)) {
          if (PrintCodeCacheExtension) {
            tty->print_cr("Extension of %s failed. Trying to allocate in %s.",
                          heap->name(), get_code_heap(type)->name());
          }
          return allocate(size, type, handle_alloc_failure, orig_code_blob_type);
        }
      }
      if (handle_alloc_failure) {
        MutexUnlocker mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);
        CompileBroker::handle_full_code_cache(orig_code_blob_type);
      }
      return NULL;
    }
    if (PrintCodeCacheExtension) {
      ResourceMark rm;
      if (_nmethod_heaps->length() >= 1) {
        tty->print("%s", heap->name());
      } else {
        tty->print("CodeCache");
      }
      tty->print_cr(" extended to [" INTPTR_FORMAT ", " INTPTR_FORMAT "] (" SSIZE_FORMAT " bytes)",
                    (intptr_t)heap->low_boundary(), (intptr_t)heap->high(),
                    (address)heap->high() - (address)heap->low_boundary());
    }
  }
  print_trace("allocation", cb, size);
  return cb;
}

void CodeCache::free(CodeBlob* cb) {
  assert_locked_or_safepoint(CodeCache_lock);
  CodeHeap* heap = get_code_heap(cb);
  print_trace("free", cb);
  if (cb->is_nmethod()) {
    heap->set_nmethod_count(heap->nmethod_count() - 1);
    if (((nmethod *)cb)->has_dependencies()) {
      _number_of_nmethods_with_dependencies--;
    }
  }
  if (cb->is_adapter_blob()) {
    heap->set_adapter_count(heap->adapter_count() - 1);
  }

  // Get heap for given CodeBlob and deallocate
  get_code_heap(cb)->deallocate(cb);

  assert(heap->blob_count() >= 0, "sanity check");
}

void CodeCache::free_unused_tail(CodeBlob* cb, size_t used) {
  assert_locked_or_safepoint(CodeCache_lock);
  guarantee(cb->is_buffer_blob() && strncmp("Interpreter", cb->name(), 11) == 0, "Only possible for interpreter!");
  print_trace("free_unused_tail", cb);

  // We also have to account for the extra space (i.e. header) used by the CodeBlob
  // which provides the memory (see BufferBlob::create() in codeBlob.cpp).
  used += CodeBlob::align_code_offset(cb->header_size());

  // Get heap for given CodeBlob and deallocate its unused tail
  get_code_heap(cb)->deallocate_tail(cb, used);
  // Adjust the sizes of the CodeBlob
  cb->adjust_size(used);
}

void CodeCache::commit(CodeBlob* cb) {
  // this is called by nmethod::nmethod, which must already own CodeCache_lock
  assert_locked_or_safepoint(CodeCache_lock);
  CodeHeap* heap = get_code_heap(cb);
  if (cb->is_nmethod()) {
    heap->set_nmethod_count(heap->nmethod_count() + 1);
    if (((nmethod *)cb)->has_dependencies()) {
      _number_of_nmethods_with_dependencies++;
    }
  }
  if (cb->is_adapter_blob()) {
    heap->set_adapter_count(heap->adapter_count() + 1);
  }

  // flush the hardware I-cache
  ICache::invalidate_range(cb->content_begin(), cb->content_size());
}

bool CodeCache::contains(void *p) {
  // S390 uses contains() in current_frame(), which is used before
  // code cache initialization if NativeMemoryTracking=detail is set.
  S390_ONLY(if (_heaps == NULL) return false;)
  // It should be ok to call contains without holding a lock.
  FOR_ALL_HEAPS(heap) {
    if ((*heap)->contains(p)) {
      return true;
    }
  }
  return false;
}

bool CodeCache::contains(nmethod *nm) {
  return contains((void *)nm);
}

// This method is safe to call without holding the CodeCache_lock. It only depends on the _segmap to contain
// valid indices, which it will always do, as long as the CodeBlob is not in the process of being recycled.
CodeBlob* CodeCache::find_blob(void* start) {
  // NMT can walk the stack before code cache is created
  if (_heaps != NULL) {
    CodeHeap* heap = get_code_heap_containing(start);
    if (heap != NULL) {
      return heap->find_blob(start);
    }
  }
  return NULL;
}

nmethod* CodeCache::find_nmethod(void* start) {
  CodeBlob* cb = find_blob(start);
  assert(cb->is_nmethod(), "did not find an nmethod");
  return (nmethod*)cb;
}

void CodeCache::blobs_do(void f(CodeBlob* nm)) {
  assert_locked_or_safepoint(CodeCache_lock);
  FOR_ALL_HEAPS(heap) {
    FOR_ALL_BLOBS(cb, *heap) {
      f(cb);
    }
  }
}

void CodeCache::nmethods_do(void f(nmethod* nm)) {
  assert_locked_or_safepoint(CodeCache_lock);
  NMethodIterator iter(NMethodIterator::all_blobs);
  while(iter.next()) {
    f(iter.method());
  }
}

void CodeCache::metadata_do(MetadataClosure* f) {
  assert_locked_or_safepoint(CodeCache_lock);
  NMethodIterator iter(NMethodIterator::all_blobs);
  while(iter.next()) {
    iter.method()->metadata_do(f);
  }
}

int CodeCache::alignment_unit() {
  return (int)_heaps->first()->alignment_unit();
}

int CodeCache::alignment_offset() {
  return (int)_heaps->first()->alignment_offset();
}

// Calculate the number of GCs after which an nmethod is expected to have been
// used in order to not be classed as cold.
void CodeCache::update_cold_gc_count() {
  if (!MethodFlushing || !UseCodeCacheFlushing || NmethodSweepActivity == 0) {
    // No aging
    return;
  }

  size_t last_used = _last_unloading_used;
  double last_time = _last_unloading_time;

  double time = os::elapsedTime();

  size_t free = unallocated_capacity();
  size_t max = max_capacity();
  size_t used = max - free;
  double gc_interval = time - last_time;

  _unloading_threshold_gc_requested = false;
  _last_unloading_time = time;
  _last_unloading_used = used;

  if (last_time == 0.0) {
    // The first GC doesn't have enough information to make good
    // decisions, so just keep everything afloat
    log_info(codecache)("Unknown code cache pressure; don't age code");
    return;
  }

  if (gc_interval <= 0.0 || last_used >= used) {
    // Dodge corner cases where there is no pressure or negative pressure
    // on the code cache. Just don't unload when this happens.
    _cold_gc_count = INT_MAX;
    log_info(codecache)("No code cache pressure; don't age code");
    return;
  }

  double allocation_rate = (used - last_used) / gc_interval;

  _unloading_allocation_rates.add(allocation_rate);
  _unloading_gc_intervals.add(gc_interval);

  size_t aggressive_sweeping_free_threshold = StartAggressiveSweepingAt / 100.0 * max;
  if (free < aggressive_sweeping_free_threshold) {
    // We are already in the red zone; be very aggressive to avoid disaster
    // But not more aggressive than 2. This ensures that an nmethod must
    // have been unused at least between two GCs to be considered cold still.
    _cold_gc_count = 2;
    log_info(codecache)("Code cache critically low; use aggressive aging");
    return;
  }

  // The code cache has an expected time for cold nmethods to "time out"
  // when they have not been used. The time for nmethods to time out
  // depends on how long we expect we can keep allocating code until
  // aggressive sweeping starts, based on sampled allocation rates.
  double average_gc_interval = _unloading_gc_intervals.avg();
  double average_allocation_rate = _unloading_allocation_rates.avg();
  double time_to_aggressive = ((double)(free - aggressive_sweeping_free_threshold)) / average_allocation_rate;
  double cold_timeout = time_to_aggressive / NmethodSweepActivity;

  // Convert time to GC cycles, and crop at INT_MAX. The reason for
  // that is that the _cold_gc_count will be added to an epoch number
  // and that addition must not overflow, or we can crash the VM.
  // But not more aggressive than 2. This ensures that an nmethod must
  // have been unused at least between two GCs to be considered cold still.
  _cold_gc_count = MAX2(MIN2((uint64_t)(cold_timeout / average_gc_interval), (uint64_t)INT_MAX), (uint64_t)2);

  double used_ratio = double(used) / double(max);
  double last_used_ratio = double(last_used) / double(max);
  log_info(codecache)("Allocation rate: %.3f KB/s, time to aggressive unloading: %.3f s, cold timeout: %.3f s, cold gc count: " UINT64_FORMAT
                      ", used: %.3f MB (%.3f%%), last used: %.3f MB (%.3f%%), gc interval: %.3f s",
                      average_allocation_rate / K, time_to_aggressive, cold_timeout, _cold_gc_count,
                      double(used) / M, used_ratio * 100.0, double(last_used) / M, last_used_ratio * 100.0, average_gc_interval);

}

uint64_t CodeCache::cold_gc_count() {
  return _cold_gc_count;
}

void CodeCache::gc_on_allocation() {
  if (!is_init_completed()) {
    // Let's not heuristically trigger GCs before the JVM is ready for GCs, no matter what
    return;
  }

  size_t free = unallocated_capacity();
  size_t max = max_capacity();
  size_t used = max - free;
  double free_ratio = double(free) / double(max);
  if (free_ratio <= StartAggressiveSweepingAt / 100.0)  {
    // In case the GC is concurrent, we make sure only one thread requests the GC.
    if (Atomic::cmpxchg(&_unloading_threshold_gc_requested, false, true) == false) {
      log_info(codecache)("Triggering aggressive GC due to having only %.3f%% free memory", free_ratio * 100.0);
      Universe::heap()->collect(GCCause::_codecache_GC_aggressive);
    }
    return;
  }

  size_t last_used = _last_unloading_used;
  if (last_used >= used) {
    // No increase since last GC; no need to sweep yet
    return;
  }
  size_t allocated_since_last = used - last_used;
  double allocated_since_last_ratio = double(allocated_since_last) / double(max);
  double threshold = SweeperThreshold / 100.0;
  double used_ratio = double(used) / double(max);
  double last_used_ratio = double(last_used) / double(max);
  if (used_ratio > threshold) {
    // After threshold is reached, scale it by free_ratio so that more aggressive
    // GC is triggered as we approach code cache exhaustion
    threshold *= free_ratio;
  }
  // If code cache has been allocated without any GC at all, let's make sure
  // it is eventually invoked to avoid trouble.
  if (allocated_since_last_ratio > threshold) {
    // In case the GC is concurrent, we make sure only one thread requests the GC.
    if (Atomic::cmpxchg(&_unloading_threshold_gc_requested, false, true) == false) {
      log_info(codecache)("Triggering threshold (%.3f%%) GC due to allocating %.3f%% since last unloading (%.3f%% used -> %.3f%% used)",
                          threshold * 100.0, allocated_since_last_ratio * 100.0, last_used_ratio * 100.0, used_ratio * 100.0);
      Universe::heap()->collect(GCCause::_codecache_GC_threshold);
    }
  }
}

// We initialize the _gc_epoch to 2, because previous_completed_gc_marking_cycle
// subtracts the value by 2, and the type is unsigned. We don't want underflow.
//
// Odd values mean that marking is in progress, and even values mean that no
// marking is currently active.
uint64_t CodeCache::_gc_epoch = 2;

// How many GCs after an nmethod has not been used, do we consider it cold?
uint64_t CodeCache::_cold_gc_count = INT_MAX;

double CodeCache::_last_unloading_time = 0.0;
size_t CodeCache::_last_unloading_used = 0;
volatile bool CodeCache::_unloading_threshold_gc_requested = false;
TruncatedSeq CodeCache::_unloading_gc_intervals(10 /* samples */);
TruncatedSeq CodeCache::_unloading_allocation_rates(10 /* samples */);

uint64_t CodeCache::gc_epoch() {
  return _gc_epoch;
}

bool CodeCache::is_gc_marking_cycle_active() {
  // Odd means that marking is active
  return (_gc_epoch % 2) == 1;
}

uint64_t CodeCache::previous_completed_gc_marking_cycle() {
  if (is_gc_marking_cycle_active()) {
    return _gc_epoch - 2;
  } else {
    return _gc_epoch - 1;
  }
}

void CodeCache::on_gc_marking_cycle_start() {
  assert(!is_gc_marking_cycle_active(), "Previous marking cycle never ended");
  ++_gc_epoch;
}

void CodeCache::on_gc_marking_cycle_finish() {
  assert(is_gc_marking_cycle_active(), "Marking cycle started before last one finished");
  ++_gc_epoch;
  update_cold_gc_count();
}

void CodeCache::arm_all_nmethods() {
  BarrierSetNMethod* bs_nm = BarrierSet::barrier_set()->barrier_set_nmethod();
  if (bs_nm != NULL) {
    bs_nm->arm_all_nmethods();
  }
}

// Mark nmethods for unloading if they contain otherwise unreachable oops.
void CodeCache::do_unloading(bool unloading_occurred) {
  assert_locked_or_safepoint(CodeCache_lock);
  CompiledMethodIterator iter(CompiledMethodIterator::all_blobs);
  while(iter.next()) {
    iter.method()->do_unloading(unloading_occurred);
  }
}

void CodeCache::blobs_do(CodeBlobClosure* f) {
  assert_locked_or_safepoint(CodeCache_lock);
  FOR_ALL_ALLOCABLE_HEAPS(heap) {
    FOR_ALL_BLOBS(cb, *heap) {
      f->do_code_blob(cb);
#ifdef ASSERT
      if (cb->is_nmethod()) {
        Universe::heap()->verify_nmethod((nmethod*)cb);
      }
#endif //ASSERT
    }
  }
}

void CodeCache::verify_clean_inline_caches() {
#ifdef ASSERT
  NMethodIterator iter(NMethodIterator::only_not_unloading);
  while(iter.next()) {
    nmethod* nm = iter.method();
    nm->verify_clean_inline_caches();
    nm->verify();
  }
#endif
}

void CodeCache::verify_icholder_relocations() {
#ifdef ASSERT
  // make sure that we aren't leaking icholders
  int count = 0;
  FOR_ALL_HEAPS(heap) {
    FOR_ALL_BLOBS(cb, *heap) {
      CompiledMethod *nm = cb->as_compiled_method_or_null();
      if (nm != NULL) {
        count += nm->verify_icholder_relocations();
      }
    }
  }
  assert(count + InlineCacheBuffer::pending_icholder_count() + CompiledICHolder::live_not_claimed_count() ==
         CompiledICHolder::live_count(), "must agree");
#endif
}

// Defer freeing of concurrently cleaned ExceptionCache entries until
// after a global handshake operation.
void CodeCache::release_exception_cache(ExceptionCache* entry) {
  if (SafepointSynchronize::is_at_safepoint()) {
    delete entry;
  } else {
    for (;;) {
      ExceptionCache* purge_list_head = Atomic::load(&_exception_cache_purge_list);
      entry->set_purge_list_next(purge_list_head);
      if (Atomic::cmpxchg(&_exception_cache_purge_list, purge_list_head, entry) == purge_list_head) {
        break;
      }
    }
  }
}

// Delete exception caches that have been concurrently unlinked,
// followed by a global handshake operation.
void CodeCache::purge_exception_caches() {
  ExceptionCache* curr = _exception_cache_purge_list;
  while (curr != NULL) {
    ExceptionCache* next = curr->purge_list_next();
    delete curr;
    curr = next;
  }
  _exception_cache_purge_list = NULL;
}

// Register an is_unloading nmethod to be flushed after unlinking
void CodeCache::register_unlinked(nmethod* nm) {
  assert(nm->unlinked_next() == NULL, "Only register for unloading once");
  for (;;) {
    // Only need acquire when reading the head, when the next
    // pointer is walked, which it is not here.
    nmethod* head = Atomic::load(&_unlinked_head);
    nmethod* next = head != NULL ? head : nm; // Self looped means end of list
    nm->set_unlinked_next(next);
    if (Atomic::cmpxchg(&_unlinked_head, head, nm) == head) {
      break;
    }
  }
}

// Flush all the nmethods the GC unlinked
void CodeCache::flush_unlinked_nmethods() {
  nmethod* nm = _unlinked_head;
  _unlinked_head = NULL;
  size_t freed_memory = 0;
  while (nm != NULL) {
    nmethod* next = nm->unlinked_next();
    freed_memory += nm->total_size();
    nm->flush();
    if (next == nm) {
      // Self looped means end of list
      break;
    }
    nm = next;
  }

  // Try to start the compiler again if we freed any memory
  if (!CompileBroker::should_compile_new_jobs() && freed_memory != 0) {
    CompileBroker::set_should_compile_new_jobs(CompileBroker::run_compilation);
    log_info(codecache)("Restarting compiler");
    EventJITRestart event;
    event.set_freedMemory(freed_memory);
    event.set_codeCacheMaxCapacity(CodeCache::max_capacity());
    event.commit();
  }
}

uint8_t CodeCache::_unloading_cycle = 1;
nmethod* volatile CodeCache::_unlinked_head = NULL;

void CodeCache::increment_unloading_cycle() {
  // 2-bit value (see IsUnloadingState in nmethod.cpp for details)
  // 0 is reserved for new methods.
  _unloading_cycle = (_unloading_cycle + 1) % 4;
  if (_unloading_cycle == 0) {
    _unloading_cycle = 1;
  }
}

CodeCache::UnloadingScope::UnloadingScope(BoolObjectClosure* is_alive)
  : _is_unloading_behaviour(is_alive)
{
  _saved_behaviour = IsUnloadingBehaviour::current();
  IsUnloadingBehaviour::set_current(&_is_unloading_behaviour);
  increment_unloading_cycle();
  DependencyContext::cleaning_start();
}

CodeCache::UnloadingScope::~UnloadingScope() {
  IsUnloadingBehaviour::set_current(_saved_behaviour);
  DependencyContext::cleaning_end();
  CodeCache::flush_unlinked_nmethods();
}

void CodeCache::verify_oops() {
  MutexLocker mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);
  VerifyOopClosure voc;
  NMethodIterator iter(NMethodIterator::only_not_unloading);
  while(iter.next()) {
    nmethod* nm = iter.method();
    nm->oops_do(&voc);
    nm->verify_oop_relocations();
  }
}

int CodeCache::blob_count(CodeBlobType code_blob_type) {
  CodeHeap* heap = get_code_heap(code_blob_type);
  return (heap != NULL) ? heap->blob_count() : 0;
}

int CodeCache::blob_count() {
  int count = 0;
  FOR_ALL_HEAPS(heap) {
    count += (*heap)->blob_count();
  }
  return count;
}

int CodeCache::nmethod_count(CodeBlobType code_blob_type) {
  CodeHeap* heap = get_code_heap(code_blob_type);
  return (heap != NULL) ? heap->nmethod_count() : 0;
}

int CodeCache::nmethod_count() {
  int count = 0;
  FOR_ALL_NMETHOD_HEAPS(heap) {
    count += (*heap)->nmethod_count();
  }
  return count;
}

int CodeCache::adapter_count(CodeBlobType code_blob_type) {
  CodeHeap* heap = get_code_heap(code_blob_type);
  return (heap != NULL) ? heap->adapter_count() : 0;
}

int CodeCache::adapter_count() {
  int count = 0;
  FOR_ALL_HEAPS(heap) {
    count += (*heap)->adapter_count();
  }
  return count;
}

address CodeCache::low_bound(CodeBlobType code_blob_type) {
  CodeHeap* heap = get_code_heap(code_blob_type);
  return (heap != NULL) ? (address)heap->low_boundary() : NULL;
}

address CodeCache::high_bound(CodeBlobType code_blob_type) {
  CodeHeap* heap = get_code_heap(code_blob_type);
  return (heap != NULL) ? (address)heap->high_boundary() : NULL;
}

size_t CodeCache::capacity() {
  size_t cap = 0;
  FOR_ALL_ALLOCABLE_HEAPS(heap) {
    cap += (*heap)->capacity();
  }
  return cap;
}

size_t CodeCache::unallocated_capacity(CodeBlobType code_blob_type) {
  CodeHeap* heap = get_code_heap(code_blob_type);
  return (heap != NULL) ? heap->unallocated_capacity() : 0;
}

size_t CodeCache::unallocated_capacity() {
  size_t unallocated_cap = 0;
  FOR_ALL_ALLOCABLE_HEAPS(heap) {
    unallocated_cap += (*heap)->unallocated_capacity();
  }
  return unallocated_cap;
}

size_t CodeCache::max_capacity() {
  size_t max_cap = 0;
  FOR_ALL_ALLOCABLE_HEAPS(heap) {
    max_cap += (*heap)->max_capacity();
  }
  return max_cap;
}

bool CodeCache::is_non_nmethod(address addr) {
  CodeHeap* blob = get_code_heap(CodeBlobType::NonNMethod);
  return blob->contains(addr);
}

size_t CodeCache::max_distance_to_non_nmethod() {
  if (!SegmentedCodeCache) {
    return ReservedCodeCacheSize;
  } else {
    CodeHeap* blob = get_code_heap(CodeBlobType::NonNMethod);
    // the max distance is minimized by placing the NonNMethod segment
    // in between MethodProfiled and MethodNonProfiled segments
    size_t dist1 = (size_t)blob->high() - (size_t)_low_bound;
    size_t dist2 = (size_t)_high_bound - (size_t)blob->low();
    return dist1 > dist2 ? dist1 : dist2;
  }
}

// Returns the reverse free ratio. E.g., if 25% (1/4) of the code cache
// is free, reverse_free_ratio() returns 4.
// Since code heap for each type of code blobs falls forward to the next
// type of code heap, return the reverse free ratio for the entire
// code cache.
double CodeCache::reverse_free_ratio() {
  double unallocated = MAX2((double)unallocated_capacity(), 1.0); // Avoid division by 0;
  double max = (double)max_capacity();
  double result = max / unallocated;
  assert (max >= unallocated, "Must be");
  assert (result >= 1.0, "reverse_free_ratio must be at least 1. It is %f", result);
  return result;
}

size_t CodeCache::bytes_allocated_in_freelists() {
  size_t allocated_bytes = 0;
  FOR_ALL_ALLOCABLE_HEAPS(heap) {
    allocated_bytes += (*heap)->allocated_in_freelist();
  }
  return allocated_bytes;
}

int CodeCache::allocated_segments() {
  int number_of_segments = 0;
  FOR_ALL_ALLOCABLE_HEAPS(heap) {
    number_of_segments += (*heap)->allocated_segments();
  }
  return number_of_segments;
}

size_t CodeCache::freelists_length() {
  size_t length = 0;
  FOR_ALL_ALLOCABLE_HEAPS(heap) {
    length += (*heap)->freelist_length();
  }
  return length;
}

void icache_init();

void CodeCache::initialize() {
  assert(CodeCacheSegmentSize >= (uintx)CodeEntryAlignment, "CodeCacheSegmentSize must be large enough to align entry points");
#ifdef COMPILER2
  assert(CodeCacheSegmentSize >= (uintx)OptoLoopAlignment,  "CodeCacheSegmentSize must be large enough to align inner loops");
#endif
  assert(CodeCacheSegmentSize >= sizeof(jdouble),    "CodeCacheSegmentSize must be large enough to align constants");
  // This was originally just a check of the alignment, causing failure, instead, round
  // the code cache to the page size.  In particular, Solaris is moving to a larger
  // default page size.
  CodeCacheExpansionSize = align_up(CodeCacheExpansionSize, os::vm_page_size());

  if (SegmentedCodeCache) {
    // Use multiple code heaps
    initialize_heaps();
  } else {
    // Use a single code heap
    FLAG_SET_ERGO(NonNMethodCodeHeapSize, 0);
    FLAG_SET_ERGO(ProfiledCodeHeapSize, 0);
    FLAG_SET_ERGO(NonProfiledCodeHeapSize, 0);
    ReservedCodeSpace rs = reserve_heap_memory(ReservedCodeCacheSize);
    add_heap(rs, "CodeCache", CodeBlobType::All);
  }

  // Initialize ICache flush mechanism
  // This service is needed for os::register_code_area
  icache_init();

  // Give OS a chance to register generated code area.
  // This is used on Windows 64 bit platforms to register
  // Structured Exception Handlers for our generated code.
  os::register_code_area((char*)low_bound(), (char*)high_bound());
}

void codeCache_init() {
  CodeCache::initialize();
}

//------------------------------------------------------------------------------------------------

int CodeCache::number_of_nmethods_with_dependencies() {
  return _number_of_nmethods_with_dependencies;
}

void CodeCache::clear_inline_caches() {
  assert_locked_or_safepoint(CodeCache_lock);
  CompiledMethodIterator iter(CompiledMethodIterator::only_not_unloading);
  while(iter.next()) {
    iter.method()->clear_inline_caches();
  }
}

// Only used by whitebox API
void CodeCache::cleanup_inline_caches_whitebox() {
  assert_locked_or_safepoint(CodeCache_lock);
  NMethodIterator iter(NMethodIterator::only_not_unloading);
  while(iter.next()) {
    iter.method()->cleanup_inline_caches_whitebox();
  }
}

// Keeps track of time spent for checking dependencies
NOT_PRODUCT(static elapsedTimer dependentCheckTime;)

int CodeCache::mark_for_deoptimization(KlassDepChange& changes) {
  MutexLocker mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);
  int number_of_marked_CodeBlobs = 0;

  // search the hierarchy looking for nmethods which are affected by the loading of this class

  // then search the interfaces this class implements looking for nmethods
  // which might be dependent of the fact that an interface only had one
  // implementor.
  // nmethod::check_all_dependencies works only correctly, if no safepoint
  // can happen
  NoSafepointVerifier nsv;
  for (DepChange::ContextStream str(changes, nsv); str.next(); ) {
    Klass* d = str.klass();
    number_of_marked_CodeBlobs += InstanceKlass::cast(d)->mark_dependent_nmethods(changes);
  }

#ifndef PRODUCT
  if (VerifyDependencies) {
    // Object pointers are used as unique identifiers for dependency arguments. This
    // is only possible if no safepoint, i.e., GC occurs during the verification code.
    dependentCheckTime.start();
    nmethod::check_all_dependencies(changes);
    dependentCheckTime.stop();
  }
#endif

  return number_of_marked_CodeBlobs;
}

CompiledMethod* CodeCache::find_compiled(void* start) {
  CodeBlob *cb = find_blob(start);
  assert(cb == NULL || cb->is_compiled(), "did not find an compiled_method");
  return (CompiledMethod*)cb;
}

#if INCLUDE_JVMTI
// RedefineClasses support for saving nmethods that are dependent on "old" methods.
// We don't really expect this table to grow very large.  If it does, it can become a hashtable.
static GrowableArray<CompiledMethod*>* old_compiled_method_table = NULL;

static void add_to_old_table(CompiledMethod* c) {
  if (old_compiled_method_table == NULL) {
    old_compiled_method_table = new (mtCode) GrowableArray<CompiledMethod*>(100, mtCode);
  }
  old_compiled_method_table->push(c);
}

static void reset_old_method_table() {
  if (old_compiled_method_table != NULL) {
    delete old_compiled_method_table;
    old_compiled_method_table = NULL;
  }
}

// Remove this method when flushed.
void CodeCache::unregister_old_nmethod(CompiledMethod* c) {
  assert_lock_strong(CodeCache_lock);
  if (old_compiled_method_table != NULL) {
    int index = old_compiled_method_table->find(c);
    if (index != -1) {
      old_compiled_method_table->delete_at(index);
    }
  }
}

void CodeCache::old_nmethods_do(MetadataClosure* f) {
  // Walk old method table and mark those on stack.
  int length = 0;
  if (old_compiled_method_table != NULL) {
    length = old_compiled_method_table->length();
    for (int i = 0; i < length; i++) {
      // Walk all methods saved on the last pass.  Concurrent class unloading may
      // also be looking at this method's metadata, so don't delete it yet if
      // it is marked as unloaded.
      old_compiled_method_table->at(i)->metadata_do(f);
    }
  }
  log_debug(redefine, class, nmethod)("Walked %d nmethods for mark_on_stack", length);
}

// Walk compiled methods and mark dependent methods for deoptimization.
int CodeCache::mark_dependents_for_evol_deoptimization() {
  assert(SafepointSynchronize::is_at_safepoint(), "Can only do this at a safepoint!");
  // Each redefinition creates a new set of nmethods that have references to "old" Methods
  // So delete old method table and create a new one.
  reset_old_method_table();

  int number_of_marked_CodeBlobs = 0;
  CompiledMethodIterator iter(CompiledMethodIterator::all_blobs);
  while(iter.next()) {
    CompiledMethod* nm = iter.method();
    // Walk all alive nmethods to check for old Methods.
    // This includes methods whose inline caches point to old methods, so
    // inline cache clearing is unnecessary.
    if (nm->has_evol_metadata()) {
      nm->mark_for_deoptimization();
      add_to_old_table(nm);
      number_of_marked_CodeBlobs++;
    }
  }

  // return total count of nmethods marked for deoptimization, if zero the caller
  // can skip deoptimization
  return number_of_marked_CodeBlobs;
}

void CodeCache::mark_all_nmethods_for_evol_deoptimization() {
  assert(SafepointSynchronize::is_at_safepoint(), "Can only do this at a safepoint!");
  CompiledMethodIterator iter(CompiledMethodIterator::all_blobs);
  while(iter.next()) {
    CompiledMethod* nm = iter.method();
    if (!nm->method()->is_method_handle_intrinsic()) {
      if (nm->can_be_deoptimized()) {
        nm->mark_for_deoptimization();
      }
      if (nm->has_evol_metadata()) {
        add_to_old_table(nm);
      }
    }
  }
}

// Flushes compiled methods dependent on redefined classes, that have already been
// marked for deoptimization.
void CodeCache::flush_evol_dependents() {
  assert(SafepointSynchronize::is_at_safepoint(), "Can only do this at a safepoint!");

  // CodeCache can only be updated by a thread_in_VM and they will all be
  // stopped during the safepoint so CodeCache will be safe to update without
  // holding the CodeCache_lock.

  // At least one nmethod has been marked for deoptimization

  Deoptimization::deoptimize_all_marked();
}
#endif // INCLUDE_JVMTI

// Mark methods for deopt (if safe or possible).
void CodeCache::mark_all_nmethods_for_deoptimization() {
  MutexLocker mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);
  CompiledMethodIterator iter(CompiledMethodIterator::only_not_unloading);
  while(iter.next()) {
    CompiledMethod* nm = iter.method();
    if (!nm->is_native_method()) {
      nm->mark_for_deoptimization();
    }
  }
}

int CodeCache::mark_for_deoptimization(Method* dependee) {
  MutexLocker mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);
  int number_of_marked_CodeBlobs = 0;

  CompiledMethodIterator iter(CompiledMethodIterator::only_not_unloading);
  while(iter.next()) {
    CompiledMethod* nm = iter.method();
    if (nm->is_dependent_on_method(dependee)) {
      ResourceMark rm;
      nm->mark_for_deoptimization();
      number_of_marked_CodeBlobs++;
    }
  }

  return number_of_marked_CodeBlobs;
}

void CodeCache::make_marked_nmethods_deoptimized() {
  RelaxedCompiledMethodIterator iter(RelaxedCompiledMethodIterator::only_not_unloading);
  while(iter.next()) {
    CompiledMethod* nm = iter.method();
    if (nm->is_marked_for_deoptimization() && !nm->has_been_deoptimized() && nm->can_be_deoptimized()) {
      nm->make_not_entrant();
      make_nmethod_deoptimized(nm);
    }
  }
}

void CodeCache::make_nmethod_deoptimized(CompiledMethod* nm) {
  if (nm->is_marked_for_deoptimization() && nm->can_be_deoptimized()) {
    nm->make_deoptimized();
  }
}

// Flushes compiled methods dependent on dependee.
void CodeCache::flush_dependents_on(InstanceKlass* dependee) {
  assert_lock_strong(Compile_lock);

  if (number_of_nmethods_with_dependencies() == 0) return;

  int marked = 0;
  if (dependee->is_linked()) {
    // Class initialization state change.
    KlassInitDepChange changes(dependee);
    marked = mark_for_deoptimization(changes);
  } else {
    // New class is loaded.
    NewKlassDepChange changes(dependee);
    marked = mark_for_deoptimization(changes);
  }

  if (marked > 0) {
    // At least one nmethod has been marked for deoptimization
    Deoptimization::deoptimize_all_marked();
  }
}

// Flushes compiled methods dependent on dependee
void CodeCache::flush_dependents_on_method(const methodHandle& m_h) {
  // --- Compile_lock is not held. However we are at a safepoint.
  assert_locked_or_safepoint(Compile_lock);

  // Compute the dependent nmethods
  if (mark_for_deoptimization(m_h()) > 0) {
    Deoptimization::deoptimize_all_marked();
  }
}

void CodeCache::verify() {
  assert_locked_or_safepoint(CodeCache_lock);
  FOR_ALL_HEAPS(heap) {
    (*heap)->verify();
    FOR_ALL_BLOBS(cb, *heap) {
      cb->verify();
    }
  }
}

// A CodeHeap is full. Print out warning and report event.
PRAGMA_DIAG_PUSH
PRAGMA_FORMAT_NONLITERAL_IGNORED
void CodeCache::report_codemem_full(CodeBlobType code_blob_type, bool print) {
  // Get nmethod heap for the given CodeBlobType and build CodeCacheFull event
  CodeHeap* heap = get_code_heap(code_blob_type);
  assert(heap != NULL, "heap is null");

  int full_count = heap->report_full();

  if ((full_count == 1) || print) {
    // Not yet reported for this heap, report
    if (SegmentedCodeCache) {
      ResourceMark rm;
      stringStream msg1_stream, msg2_stream;
      msg1_stream.print("%s is full. Compiler has been disabled.",
                        get_code_heap_name(code_blob_type));
      msg2_stream.print("Try increasing the code heap size using -XX:%s=",
                 get_code_heap_flag_name(code_blob_type));
      const char *msg1 = msg1_stream.as_string();
      const char *msg2 = msg2_stream.as_string();

      log_warning(codecache)("%s", msg1);
      log_warning(codecache)("%s", msg2);
      warning("%s", msg1);
      warning("%s", msg2);
    } else {
      const char *msg1 = "CodeCache is full. Compiler has been disabled.";
      const char *msg2 = "Try increasing the code cache size using -XX:ReservedCodeCacheSize=";

      log_warning(codecache)("%s", msg1);
      log_warning(codecache)("%s", msg2);
      warning("%s", msg1);
      warning("%s", msg2);
    }
    stringStream s;
    // Dump code cache into a buffer before locking the tty.
    {
      MutexLocker mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);
      print_summary(&s);
    }
    {
      ttyLocker ttyl;
      tty->print("%s", s.freeze());
    }

    if (full_count == 1) {
      if (PrintCodeHeapAnalytics) {
        CompileBroker::print_heapinfo(tty, "all", 4096); // details, may be a lot!
      }
    }
  }

  EventCodeCacheFull event;
  if (event.should_commit()) {
    event.set_codeBlobType((u1)code_blob_type);
    event.set_startAddress((u8)heap->low_boundary());
    event.set_commitedTopAddress((u8)heap->high());
    event.set_reservedTopAddress((u8)heap->high_boundary());
    event.set_entryCount(heap->blob_count());
    event.set_methodCount(heap->nmethod_count());
    event.set_adaptorCount(heap->adapter_count());
    event.set_unallocatedCapacity(heap->unallocated_capacity());
    event.set_fullCount(heap->full_count());
    event.set_codeCacheMaxCapacity(CodeCache::max_capacity());
    event.commit();
  }
}
PRAGMA_DIAG_POP

void CodeCache::print_memory_overhead() {
  size_t wasted_bytes = 0;
  FOR_ALL_ALLOCABLE_HEAPS(heap) {
      CodeHeap* curr_heap = *heap;
      for (CodeBlob* cb = (CodeBlob*)curr_heap->first(); cb != NULL; cb = (CodeBlob*)curr_heap->next(cb)) {
        HeapBlock* heap_block = ((HeapBlock*)cb) - 1;
        wasted_bytes += heap_block->length() * CodeCacheSegmentSize - cb->size();
      }
  }
  // Print bytes that are allocated in the freelist
  ttyLocker ttl;
  tty->print_cr("Number of elements in freelist: " SSIZE_FORMAT,       freelists_length());
  tty->print_cr("Allocated in freelist: " SSIZE_FORMAT "kB",  bytes_allocated_in_freelists()/K);
  tty->print_cr("Unused bytes in CodeBlobs: " SSIZE_FORMAT "kB",  (wasted_bytes/K));
  tty->print_cr("Segment map size: " SSIZE_FORMAT "kB",  allocated_segments()/K); // 1 byte per segment
}

//------------------------------------------------------------------------------------------------
// Non-product version

#ifndef PRODUCT

void CodeCache::print_trace(const char* event, CodeBlob* cb, int size) {
  if (PrintCodeCache2) {  // Need to add a new flag
    ResourceMark rm;
    if (size == 0)  size = cb->size();
    tty->print_cr("CodeCache %s: addr: " INTPTR_FORMAT ", size: 0x%x", event, p2i(cb), size);
  }
}

void CodeCache::print_internals() {
  int nmethodCount = 0;
  int runtimeStubCount = 0;
  int adapterCount = 0;
  int deoptimizationStubCount = 0;
  int uncommonTrapStubCount = 0;
  int bufferBlobCount = 0;
  int total = 0;
  int nmethodNotEntrant = 0;
  int nmethodJava = 0;
  int nmethodNative = 0;
  int max_nm_size = 0;
  ResourceMark rm;

  int i = 0;
  FOR_ALL_ALLOCABLE_HEAPS(heap) {
    if ((_nmethod_heaps->length() >= 1) && Verbose) {
      tty->print_cr("-- %s --", (*heap)->name());
    }
    FOR_ALL_BLOBS(cb, *heap) {
      total++;
      if (cb->is_nmethod()) {
        nmethod* nm = (nmethod*)cb;

        if (Verbose && nm->method() != NULL) {
          ResourceMark rm;
          char *method_name = nm->method()->name_and_sig_as_C_string();
          tty->print("%s", method_name);
          if(nm->is_not_entrant()) { tty->print_cr(" not-entrant"); }
        }

        nmethodCount++;

        if(nm->is_not_entrant()) { nmethodNotEntrant++; }
        if(nm->method() != NULL && nm->is_native_method()) { nmethodNative++; }

        if(nm->method() != NULL && nm->is_java_method()) {
          nmethodJava++;
          max_nm_size = MAX2(max_nm_size, nm->size());
        }
      } else if (cb->is_runtime_stub()) {
        runtimeStubCount++;
      } else if (cb->is_deoptimization_stub()) {
        deoptimizationStubCount++;
      } else if (cb->is_uncommon_trap_stub()) {
        uncommonTrapStubCount++;
      } else if (cb->is_adapter_blob()) {
        adapterCount++;
      } else if (cb->is_buffer_blob()) {
        bufferBlobCount++;
      }
    }
  }

  int bucketSize = 512;
  int bucketLimit = max_nm_size / bucketSize + 1;
  int *buckets = NEW_C_HEAP_ARRAY(int, bucketLimit, mtCode);
  memset(buckets, 0, sizeof(int) * bucketLimit);

  NMethodIterator iter(NMethodIterator::all_blobs);
  while(iter.next()) {
    nmethod* nm = iter.method();
    if(nm->method() != NULL && nm->is_java_method()) {
      buckets[nm->size() / bucketSize]++;
    }
  }

  tty->print_cr("Code Cache Entries (total of %d)",total);
  tty->print_cr("-------------------------------------------------");
  tty->print_cr("nmethods: %d",nmethodCount);
  tty->print_cr("\tnot_entrant: %d",nmethodNotEntrant);
  tty->print_cr("\tjava: %d",nmethodJava);
  tty->print_cr("\tnative: %d",nmethodNative);
  tty->print_cr("runtime_stubs: %d",runtimeStubCount);
  tty->print_cr("adapters: %d",adapterCount);
  tty->print_cr("buffer blobs: %d",bufferBlobCount);
  tty->print_cr("deoptimization_stubs: %d",deoptimizationStubCount);
  tty->print_cr("uncommon_traps: %d",uncommonTrapStubCount);
  tty->print_cr("\nnmethod size distribution");
  tty->print_cr("-------------------------------------------------");

  for(int i=0; i<bucketLimit; i++) {
    if(buckets[i] != 0) {
      tty->print("%d - %d bytes",i*bucketSize,(i+1)*bucketSize);
      tty->fill_to(40);
      tty->print_cr("%d",buckets[i]);
    }
  }

  FREE_C_HEAP_ARRAY(int, buckets);
  print_memory_overhead();
}

#endif // !PRODUCT

void CodeCache::print() {
  print_summary(tty);

#ifndef PRODUCT
  if (!Verbose) return;

  CodeBlob_sizes live[CompLevel_full_optimization + 1];
  CodeBlob_sizes runtimeStub;
  CodeBlob_sizes uncommonTrapStub;
  CodeBlob_sizes deoptimizationStub;
  CodeBlob_sizes adapter;
  CodeBlob_sizes bufferBlob;
  CodeBlob_sizes other;

  FOR_ALL_ALLOCABLE_HEAPS(heap) {
    FOR_ALL_BLOBS(cb, *heap) {
      if (cb->is_nmethod()) {
        const int level = cb->as_nmethod()->comp_level();
        assert(0 <= level && level <= CompLevel_full_optimization, "Invalid compilation level");
        live[level].add(cb);
      } else if (cb->is_runtime_stub()) {
        runtimeStub.add(cb);
      } else if (cb->is_deoptimization_stub()) {
        deoptimizationStub.add(cb);
      } else if (cb->is_uncommon_trap_stub()) {
        uncommonTrapStub.add(cb);
      } else if (cb->is_adapter_blob()) {
        adapter.add(cb);
      } else if (cb->is_buffer_blob()) {
        bufferBlob.add(cb);
      } else {
        other.add(cb);
      }
    }
  }

  tty->print_cr("nmethod dependency checking time %fs", dependentCheckTime.seconds());

  tty->print_cr("nmethod blobs per compilation level:");
  for (int i = 0; i <= CompLevel_full_optimization; i++) {
    const char *level_name;
    switch (i) {
    case CompLevel_none:              level_name = "none";              break;
    case CompLevel_simple:            level_name = "simple";            break;
    case CompLevel_limited_profile:   level_name = "limited profile";   break;
    case CompLevel_full_profile:      level_name = "full profile";      break;
    case CompLevel_full_optimization: level_name = "full optimization"; break;
    default: assert(false, "invalid compilation level");
    }
    tty->print_cr("%s:", level_name);
    live[i].print("live");
  }

  struct {
    const char* name;
    const CodeBlob_sizes* sizes;
  } non_nmethod_blobs[] = {
    { "runtime",        &runtimeStub },
    { "uncommon trap",  &uncommonTrapStub },
    { "deoptimization", &deoptimizationStub },
    { "adapter",        &adapter },
    { "buffer blob",    &bufferBlob },
    { "other",          &other },
  };
  tty->print_cr("Non-nmethod blobs:");
  for (auto& blob: non_nmethod_blobs) {
    blob.sizes->print(blob.name);
  }

  if (WizardMode) {
     // print the oop_map usage
    int code_size = 0;
    int number_of_blobs = 0;
    int number_of_oop_maps = 0;
    int map_size = 0;
    FOR_ALL_ALLOCABLE_HEAPS(heap) {
      FOR_ALL_BLOBS(cb, *heap) {
        number_of_blobs++;
        code_size += cb->code_size();
        ImmutableOopMapSet* set = cb->oop_maps();
        if (set != NULL) {
          number_of_oop_maps += set->count();
          map_size           += set->nr_of_bytes();
        }
      }
    }
    tty->print_cr("OopMaps");
    tty->print_cr(" #blobs = %d", number_of_blobs);
    tty->print_cr(" code size = %d", code_size);
    tty->print_cr(" #oop_maps = %d", number_of_oop_maps);
    tty->print_cr(" map size = %d", map_size);
  }

#endif // !PRODUCT
}

void CodeCache::print_summary(outputStream* st, bool detailed) {
  int full_count = 0;
  FOR_ALL_HEAPS(heap_iterator) {
    CodeHeap* heap = (*heap_iterator);
    size_t total = (heap->high_boundary() - heap->low_boundary());
    if (_heaps->length() >= 1) {
      st->print("%s:", heap->name());
    } else {
      st->print("CodeCache:");
    }
    st->print_cr(" size=" SIZE_FORMAT "Kb used=" SIZE_FORMAT
                 "Kb max_used=" SIZE_FORMAT "Kb free=" SIZE_FORMAT "Kb",
                 total/K, (total - heap->unallocated_capacity())/K,
                 heap->max_allocated_capacity()/K, heap->unallocated_capacity()/K);

    if (detailed) {
      st->print_cr(" bounds [" INTPTR_FORMAT ", " INTPTR_FORMAT ", " INTPTR_FORMAT "]",
                   p2i(heap->low_boundary()),
                   p2i(heap->high()),
                   p2i(heap->high_boundary()));

      full_count += get_codemem_full_count(heap->code_blob_type());
    }
  }

  if (detailed) {
    st->print_cr(" total_blobs=" UINT32_FORMAT " nmethods=" UINT32_FORMAT
                       " adapters=" UINT32_FORMAT,
                       blob_count(), nmethod_count(), adapter_count());
    st->print_cr(" compilation: %s", CompileBroker::should_compile_new_jobs() ?
                 "enabled" : Arguments::mode() == Arguments::_int ?
                 "disabled (interpreter mode)" :
                 "disabled (not enough contiguous free space left)");
    st->print_cr(" stopped_count=%d, restarted_count=%d",
                 CompileBroker::get_total_compiler_stopped_count(),
                 CompileBroker::get_total_compiler_restarted_count());
    st->print_cr(" full_count=%d", full_count);
  }
}

void CodeCache::print_codelist(outputStream* st) {
  MutexLocker mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);

  CompiledMethodIterator iter(CompiledMethodIterator::only_not_unloading);
  while (iter.next()) {
    CompiledMethod* cm = iter.method();
    ResourceMark rm;
    char* method_name = cm->method()->name_and_sig_as_C_string();
    st->print_cr("%d %d %d %s [" INTPTR_FORMAT ", " INTPTR_FORMAT " - " INTPTR_FORMAT "]",
                 cm->compile_id(), cm->comp_level(), cm->get_state(),
                 method_name,
                 (intptr_t)cm->header_begin(), (intptr_t)cm->code_begin(), (intptr_t)cm->code_end());
  }
}

void CodeCache::print_layout(outputStream* st) {
  MutexLocker mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);
  ResourceMark rm;
  print_summary(st, true);
}

void CodeCache::log_state(outputStream* st) {
  st->print(" total_blobs='" UINT32_FORMAT "' nmethods='" UINT32_FORMAT "'"
            " adapters='" UINT32_FORMAT "' free_code_cache='" SIZE_FORMAT "'",
            blob_count(), nmethod_count(), adapter_count(),
            unallocated_capacity());
}

#ifdef LINUX
void CodeCache::write_perf_map() {
  MutexLocker mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);

  // Perf expects to find the map file at /tmp/perf-<pid>.map.
  char fname[32];
  jio_snprintf(fname, sizeof(fname), "/tmp/perf-%d.map", os::current_process_id());

  fileStream fs(fname, "w");
  if (!fs.is_open()) {
    log_warning(codecache)("Failed to create %s for perf map", fname);
    return;
  }

  AllCodeBlobsIterator iter(AllCodeBlobsIterator::only_not_unloading);
  while (iter.next()) {
    CodeBlob *cb = iter.method();
    ResourceMark rm;
    const char* method_name =
      cb->is_compiled() ? cb->as_compiled_method()->method()->external_name()
--> --------------------

--> maximum size reached

--> --------------------

quality90%

¤ Dauer der Verarbeitung: 0.46 Sekunden (vorverarbeitet) ¤

Wurzel

Suchen

Beweissystem der NASA

Beweissystem Isabelle

NIST Cobol Testsuite

Cephes Mathematical Library

Wiener Entwicklungsmethode

Haftungshinweis

Die Informationen auf dieser Webseite wurden nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit, noch Qualität der bereit gestellten Informationen zugesichert.

Bemerkung:

Die farbliche Syntaxdarstellung ist noch experimentell.