src/hotspot/share/gc/parallel/psOldGen.cpp

/*
 * Copyright (c) 2001, 2025, 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 "gc/parallel/objectStartArray.inline.hpp"
#include "gc/parallel/parallelArguments.hpp"
#include "gc/parallel/parallelScavengeHeap.hpp"
#include "gc/parallel/psAdaptiveSizePolicy.hpp"
#include "gc/parallel/psCardTable.hpp"
#include "gc/parallel/psOldGen.hpp"
#include "gc/shared/cardTableBarrierSet.hpp"
#include "gc/shared/gcLocker.hpp"
#include "gc/shared/spaceDecorator.hpp"
#include "logging/log.hpp"
#include "oops/oop.inline.hpp"
#include "runtime/java.hpp"
#include "utilities/align.hpp"

PSOldGen::PSOldGen(ReservedSpace rs, size_t initial_size, size_t min_size,
                   size_t max_size):
  _min_gen_size(min_size),
  _max_gen_size(max_size)
{
  initialize(rs, initial_size, GenAlignment);
}

void PSOldGen::initialize(ReservedSpace rs, size_t initial_size, size_t alignment) {
  initialize_virtual_space(rs, initial_size, alignment);
  initialize_work();

  initialize_performance_counters();
}

void PSOldGen::initialize_virtual_space(ReservedSpace rs,
                                        size_t initial_size,
                                        size_t alignment) {

  _virtual_space = new PSVirtualSpace(rs, alignment);
  if (!_virtual_space->expand_by(initial_size)) {
    vm_exit_during_initialization("Could not reserve enough space for "
                                  "object heap");
  }
}

void PSOldGen::initialize_work() {
  MemRegion const reserved_mr = reserved();
  assert(reserved_mr.byte_size() == max_gen_size(), "invariant");

  // Card table stuff: for all committed memory
  MemRegion committed_mr((HeapWord*)virtual_space()->low(),
                         (HeapWord*)virtual_space()->high());

  if (ZapUnusedHeapArea) {
    // Mangle newly committed space immediately rather than
    // waiting for the initialization of the space even though
    // mangling is related to spaces.  Doing it here eliminates
    // the need to carry along information that a complete mangling
    // (bottom to end) needs to be done.
    SpaceMangler::mangle_region(committed_mr);
  }

  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
  PSCardTable* ct = heap->card_table();
  ct->resize_covered_region(committed_mr);

  // Verify that the start and end of this generation is the start of a card.
  // If this wasn't true, a single card could span more than one generation,
  // which would cause problems when we commit/uncommit memory, and when we
  // clear and dirty cards.
  guarantee(CardTable::is_card_aligned(reserved_mr.start()), "generation must be card aligned");
  // Check the heap layout documented at `class ParallelScavengeHeap`.
  assert(reserved_mr.end() != heap->reserved_region().end(), "invariant");
  guarantee(CardTable::is_card_aligned(reserved_mr.end()), "generation must be card aligned");

  //
  // ObjectSpace stuff
  //

  _object_space = new MutableSpace(virtual_space()->alignment());
  object_space()->initialize(committed_mr,
                             SpaceDecorator::Clear,
                             SpaceDecorator::Mangle,
                             MutableSpace::SetupPages,
                             &ParallelScavengeHeap::heap()->workers());

  // Update the start_array
  _start_array = new ObjectStartArray(reserved_mr);
  start_array()->set_covered_region(committed_mr);
}

void PSOldGen::initialize_performance_counters() {
  const char* perf_data_name = "old";
  _gen_counters = new GenerationCounters(perf_data_name, 1, 1, min_gen_size(),
                                         max_gen_size(), virtual_space()->committed_size());
  _space_counters = new SpaceCounters(perf_data_name, 0,
                                      virtual_space()->reserved_size(),
                                      _object_space, _gen_counters);
}

HeapWord* PSOldGen::expand_and_allocate(size_t word_size) {
  assert(SafepointSynchronize::is_at_safepoint(), "precondition");
  assert(Thread::current()->is_VM_thread(), "precondition");
  if (object_space()->needs_expand(word_size)) {
    expand(word_size*HeapWordSize);
  }

  // Reuse the CAS API even though this is VM thread in safepoint. This method
  // is not invoked repeatedly, so the CAS overhead should be negligible.
  return cas_allocate_noexpand(word_size);
}

size_t PSOldGen::num_iterable_blocks() const {
  return (object_space()->used_in_bytes() + IterateBlockSize - 1) / IterateBlockSize;
}

void PSOldGen::object_iterate_block(ObjectClosure* cl, size_t block_index) {
  size_t block_word_size = IterateBlockSize / HeapWordSize;
  assert((block_word_size % CardTable::card_size_in_words()) == 0,
         "To ensure fast object_start calls");

  MutableSpace *space = object_space();

  HeapWord* begin = space->bottom() + block_index * block_word_size;
  HeapWord* end = MIN2(space->top(), begin + block_word_size);

  // Get object starting at or reaching into this block.
  HeapWord* start = start_array()->object_start(begin);
  if (start < begin) {
    start += cast_to_oop(start)->size();
  }
  assert(start >= begin,
         "Object address" PTR_FORMAT " must be larger or equal to block address at " PTR_FORMAT,
         p2i(start), p2i(begin));
  // Iterate all objects until the end.
  for (HeapWord* p = start; p < end; p += cast_to_oop(p)->size()) {
    cl->do_object(cast_to_oop(p));
  }
}

bool PSOldGen::expand_for_allocate(size_t word_size) {
  assert(word_size > 0, "allocating zero words?");
  bool result = true;
  {
    MutexLocker x(PSOldGenExpand_lock);
    // Avoid "expand storms" by rechecking available space after obtaining
    // the lock, because another thread may have already made sufficient
    // space available.  If insufficient space available, that will remain
    // true until we expand, since we have the lock.  Other threads may take
    // the space we need before we can allocate it, regardless of whether we
    // expand.  That's okay, we'll just try expanding again.
    if (object_space()->needs_expand(word_size)) {
      result = expand(word_size*HeapWordSize);
    }
  }
  if (GCExpandToAllocateDelayMillis > 0) {
    os::naked_sleep(GCExpandToAllocateDelayMillis);
  }
  return result;
}

bool PSOldGen::expand(size_t bytes) {
#ifdef ASSERT
  if (!Thread::current()->is_VM_thread()) {
    assert_lock_strong(PSOldGenExpand_lock);
  }
  assert_locked_or_safepoint(Heap_lock);
  assert(bytes > 0, "precondition");
#endif
  const size_t remaining_bytes = virtual_space()->uncommitted_size();
  if (remaining_bytes == 0) {
    return false;
  }
  const size_t alignment = virtual_space()->alignment();
  size_t aligned_bytes = align_up(MIN2(bytes, remaining_bytes), alignment);
  size_t aligned_expand_bytes = align_up(MinHeapDeltaBytes, alignment);

  if (UseNUMA) {
    // With NUMA we use round-robin page allocation for the old gen. Expand by at least
    // providing a page per lgroup. Alignment is larger or equal to the page size.
    aligned_expand_bytes = MAX2(aligned_expand_bytes, alignment * os::numa_get_groups_num());
  }

  bool success = false;
  if (aligned_expand_bytes > aligned_bytes) {
    success = expand_by(aligned_expand_bytes);
  }
  if (!success) {
    success = expand_by(aligned_bytes);
  }
  if (!success) {
    success = expand_to_reserved();
  }

  return success;
}

bool PSOldGen::expand_by(size_t bytes) {
  assert(bytes > 0, "precondition");
  bool result = virtual_space()->expand_by(bytes);
  if (result) {
    if (ZapUnusedHeapArea) {
      // We need to mangle the newly expanded area. The memregion spans
      // end -> new_end, we assume that top -> end is already mangled.
      // Do the mangling before post_resize() is called because
      // the space is available for allocation after post_resize();
      HeapWord* const virtual_space_high = (HeapWord*) virtual_space()->high();
      assert(object_space()->end() < virtual_space_high,
        "Should be true before post_resize()");
      MemRegion mangle_region(object_space()->end(), virtual_space_high);
      // Note that the object space has not yet been updated to
      // coincide with the new underlying virtual space.
      SpaceMangler::mangle_region(mangle_region);
    }
    post_resize();
    if (UsePerfData) {
      _space_counters->update_capacity();
      _gen_counters->update_all(_virtual_space->committed_size());
    }
  }

  if (result) {
    size_t new_mem_size = virtual_space()->committed_size();
    size_t old_mem_size = new_mem_size - bytes;
    log_debug(gc)("Expanding %s from %zuK by %zuK to %zuK",
                  name(), old_mem_size/K, bytes/K, new_mem_size/K);
  }

  return result;
}

bool PSOldGen::expand_to_reserved() {
  bool result = false;
  const size_t remaining_bytes = virtual_space()->uncommitted_size();
  if (remaining_bytes > 0) {
    result = expand_by(remaining_bytes);
    DEBUG_ONLY(if (!result) log_warning(gc)("grow to reserve failed"));
  }
  return result;
}

void PSOldGen::shrink(size_t bytes) {
  assert(Thread::current()->is_VM_thread(), "precondition");
  assert(SafepointSynchronize::is_at_safepoint(), "precondition");
  assert(bytes > 0, "precondition");

  size_t size = align_down(bytes, virtual_space()->alignment());
  if (size > 0) {
    virtual_space()->shrink_by(bytes);
    post_resize();

    size_t new_mem_size = virtual_space()->committed_size();
    size_t old_mem_size = new_mem_size + bytes;
    log_debug(gc)("Shrinking %s from %zuK by %zuK to %zuK",
                  name(), old_mem_size/K, bytes/K, new_mem_size/K);
  }
}

void PSOldGen::complete_loaded_archive_space(MemRegion archive_space) {
  HeapWord* cur = archive_space.start();
  while (cur < archive_space.end()) {
    size_t word_size = cast_to_oop(cur)->size();
    _start_array->update_for_block(cur, cur + word_size);
    cur += word_size;
  }
}

void PSOldGen::resize(size_t desired_free_space) {
  const size_t alignment = virtual_space()->alignment();
  const size_t size_before = virtual_space()->committed_size();
  size_t new_size = used_in_bytes() + desired_free_space;
  if (new_size < used_in_bytes()) {
    // Overflowed the addition.
    new_size = max_gen_size();
  }
  // Adjust according to our min and max
  new_size = clamp(new_size, min_gen_size(), max_gen_size());

  new_size = align_up(new_size, alignment);

  const size_t current_size = capacity_in_bytes();

  log_trace(gc, ergo)("AdaptiveSizePolicy::old generation size: "
    "desired free: %zu used: %zu"
    " new size: %zu current size %zu"
    " gen limits: %zu / %zu",
    desired_free_space, used_in_bytes(), new_size, current_size,
    max_gen_size(), min_gen_size());

  if (new_size == current_size) {
    // No change requested
    return;
  }
  if (new_size > current_size) {
    size_t change_bytes = new_size - current_size;
    expand(change_bytes);
  } else {
    size_t change_bytes = current_size - new_size;
    shrink(change_bytes);
  }

  log_trace(gc, ergo)("AdaptiveSizePolicy::old generation size: collection: %d (%zu) -> (%zu) ",
                      ParallelScavengeHeap::heap()->total_collections(),
                      size_before,
                      virtual_space()->committed_size());
}

// NOTE! We need to be careful about resizing. During a GC, multiple
// allocators may be active during heap expansion. If we allow the
// heap resizing to become visible before we have correctly resized
// all heap related data structures, we may cause program failures.
void PSOldGen::post_resize() {
  // First construct a memregion representing the new size
  MemRegion new_memregion((HeapWord*)virtual_space()->low(),
    (HeapWord*)virtual_space()->high());
  size_t new_word_size = new_memregion.word_size();

  start_array()->set_covered_region(new_memregion);
  ParallelScavengeHeap::heap()->card_table()->resize_covered_region(new_memregion);

  WorkerThreads* workers = Thread::current()->is_VM_thread() ?
                      &ParallelScavengeHeap::heap()->workers() : nullptr;

  // The update of the space's end is done by this call.  As that
  // makes the new space available for concurrent allocation, this
  // must be the last step when expanding.
  object_space()->initialize(new_memregion,
                             SpaceDecorator::DontClear,
                             SpaceDecorator::DontMangle,
                             MutableSpace::SetupPages,
                             workers);

  assert(new_word_size == heap_word_size(object_space()->capacity_in_bytes()),
    "Sanity");
}

void PSOldGen::print() const { print_on(tty);}
void PSOldGen::print_on(outputStream* st) const {
  st->print("%-15s", name());
  st->print(" total %zuK, used %zuK ", capacity_in_bytes() / K, used_in_bytes() / K);
  virtual_space()->print_space_boundaries_on(st);

  StreamIndentor si(st, 1);
  object_space()->print_on(st, "object ");
}

void PSOldGen::update_counters() {
  if (UsePerfData) {
    _space_counters->update_all();
    _gen_counters->update_all(_virtual_space->committed_size());
  }
}

void PSOldGen::verify() {
  object_space()->verify();
}