/*

 * Copyright (c) 2018, 2019, Red Hat, Inc. All rights reserved.

 *

 * 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 "classfile/classLoaderData.hpp"

#include "classfile/classLoaderDataGraph.hpp"

#include "gc/shared/referenceProcessor.hpp"

#include "gc/shared/referenceProcessorPhaseTimes.hpp"

#include "gc/shared/workgroup.hpp"

#include "gc/shared/weakProcessor.inline.hpp"

#include "gc/shenandoah/shenandoahBarrierSet.hpp"

#include "gc/shenandoah/shenandoahClosures.inline.hpp"

#include "gc/shenandoah/shenandoahCodeRoots.hpp"

#include "gc/shenandoah/shenandoahCollectionSet.hpp"

#include "gc/shenandoah/shenandoahCollectorPolicy.hpp"

#include "gc/shenandoah/shenandoahFreeSet.hpp"

#include "gc/shenandoah/shenandoahForwarding.hpp"

#include "gc/shenandoah/shenandoahPhaseTimings.hpp"

#include "gc/shenandoah/shenandoahHeap.inline.hpp"

#include "gc/shenandoah/shenandoahHeapRegionSet.inline.hpp"

#include "gc/shenandoah/shenandoahHeuristics.hpp"

#include "gc/shenandoah/shenandoahMarkingContext.inline.hpp"

#include "gc/shenandoah/shenandoahOopClosures.inline.hpp"

#include "gc/shenandoah/shenandoahRootProcessor.inline.hpp"

#include "gc/shenandoah/shenandoahStringDedup.hpp"

#include "gc/shenandoah/shenandoahTaskqueue.inline.hpp"

#include "gc/shenandoah/shenandoahTimingTracker.hpp"

#include "gc/shenandoah/shenandoahTraversalGC.hpp"

#include "gc/shenandoah/shenandoahUtils.hpp"

#include "gc/shenandoah/shenandoahVerifier.hpp"

#include "memory/iterator.hpp"

#include "memory/metaspace.hpp"

#include "memory/resourceArea.hpp"

#include "memory/universe.hpp"

/**

 * NOTE: We are using the SATB buffer in thread.hpp and satbMarkQueue.hpp, however, it is not an SATB algorithm.

 * We're using the buffer as generic oop buffer to enqueue new values in concurrent oop stores, IOW, the algorithm

 * is incremental-update-based.

 *

 * NOTE on interaction with TAMS: we want to avoid traversing new objects for

 * several reasons:

 * - We will not reclaim them in this cycle anyway, because they are not in the

 *   cset

 * - It makes up for the bulk of work during final-pause

 * - It also shortens the concurrent cycle because we don't need to

 *   pointlessly traverse through newly allocated objects.

 * - As a nice side-effect, it solves the I-U termination problem (mutators

 *   cannot outrun the GC by allocating like crazy)

 * - It is an easy way to achieve MWF. What MWF does is to also enqueue the

 *   target object of stores if it's new. Treating new objects live implicitely

 *   achieves the same, but without extra barriers. I think the effect of

 *   shortened final-pause (mentioned above) is the main advantage of MWF. In

 *   particular, we will not see the head of a completely new long linked list

 *   in final-pause and end up traversing huge chunks of the heap there.

 * - We don't need to see/update the fields of new objects either, because they

 *   are either still null, or anything that's been stored into them has been

 *   evacuated+enqueued before (and will thus be treated later).

 *

 * We achieve this by setting TAMS for each region, and everything allocated

 * beyond TAMS will be 'implicitely marked'.

 *

 * Gotchas:

 * - While we want new objects to be implicitely marked, we don't want to count

 *   them alive. Otherwise the next cycle wouldn't pick them up and consider

 *   them for cset. This means that we need to protect such regions from

 *   getting accidentally thrashed at the end of traversal cycle. This is why I

 *   keep track of alloc-regions and check is_alloc_region() in the trashing

 *   code.

 * - We *need* to traverse through evacuated objects. Those objects are

 *   pre-existing, and any references in them point to interesting objects that

 *   we need to see. We also want to count them as live, because we just

 *   determined that they are alive :-) I achieve this by upping TAMS

 *   concurrently for every gclab/gc-shared alloc before publishing the

 *   evacuated object. This way, the GC threads will not consider such objects

 *   implictely marked, and traverse through them as normal.

 */

class ShenandoahTraversalSATBBufferClosure : public SATBBufferClosure {

private:

  ShenandoahObjToScanQueue* _queue;

  ShenandoahTraversalGC* _traversal_gc;

  ShenandoahHeap* const _heap;

public:

  ShenandoahTraversalSATBBufferClosure(ShenandoahObjToScanQueue* q) :

    _queue(q),

    _heap(ShenandoahHeap::heap())

 { }

  void do_buffer(void** buffer, size_t size) {

    for (size_t i = 0; i < size; ++i) {

      oop* p = (oop*) &buffer[i];

      oop obj = RawAccess<>::oop_load(p);

      shenandoah_assert_not_forwarded(p, obj);

      if (_heap->marking_context()->mark(obj)) {

        _queue->push(ShenandoahMarkTask(obj));

      }

    }

  }

};

class ShenandoahTraversalSATBThreadsClosure : public ThreadClosure {

private:

  ShenandoahTraversalSATBBufferClosure* _satb_cl;

public:

  ShenandoahTraversalSATBThreadsClosure(ShenandoahTraversalSATBBufferClosure* satb_cl) :

    _satb_cl(satb_cl) {}

  void do_thread(Thread* thread) {

    ShenandoahThreadLocalData::satb_mark_queue(thread).apply_closure_and_empty(_satb_cl);

  }

};

// Like CLDToOopClosure, but clears has_modified_oops, so that we can record modified CLDs during traversal

// and remark them later during final-traversal.

class ShenandoahMarkCLDClosure : public CLDClosure {

private:

  OopClosure* _cl;

public:

  ShenandoahMarkCLDClosure(OopClosure* cl) : _cl(cl) {}

  void do_cld(ClassLoaderData* cld) {

    cld->oops_do(_cl, true, true);

  }

};

// Like CLDToOopClosure, but only process modified CLDs

class ShenandoahRemarkCLDClosure : public CLDClosure {

private:

  OopClosure* _cl;

public:

  ShenandoahRemarkCLDClosure(OopClosure* cl) : _cl(cl) {}

  void do_cld(ClassLoaderData* cld) {

    if (cld->has_modified_oops()) {

      cld->oops_do(_cl, true, true);

    }

  }

};

class ShenandoahInitTraversalCollectionTask : public AbstractGangTask {

private:

  ShenandoahHeap* _heap;

  ShenandoahCsetCodeRootsIterator* _cset_coderoots;

  ShenandoahStringDedupRoots       _dedup_roots;

public:

    AbstractGangTask("Shenandoah Init Traversal Collection"),

    _rp(rp),

  void work(uint worker_id) {

    ShenandoahParallelWorkerSession worker_session(worker_id);

    ShenandoahEvacOOMScope oom_evac_scope;

    ShenandoahObjToScanQueueSet* queues = _heap->traversal_gc()->task_queues();

    ShenandoahObjToScanQueue* q = queues->queue(worker_id);

    bool process_refs = _heap->process_references();

    bool unload_classes = _heap->unload_classes();

    ReferenceProcessor* rp = NULL;

    if (process_refs) {

      rp = _heap->ref_processor();

    }

    // Step 1: Process ordinary GC roots.

    {

      ShenandoahTraversalClosure roots_cl(q, rp);

      ShenandoahMarkCLDClosure cld_cl(&roots_cl);

      MarkingCodeBlobClosure code_cl(&roots_cl, CodeBlobToOopClosure::FixRelocations);

      if (unload_classes) {

        _rp->strong_roots_do(worker_id, &roots_cl);

        // Need to pre-evac code roots here. Otherwise we might see from-space constants.

        ShenandoahWorkerTimings* worker_times = _heap->phase_timings()->worker_times();

        ShenandoahWorkerTimingsTracker timer(worker_times, ShenandoahPhaseTimings::CodeCacheRoots, worker_id);

        _cset_coderoots->possibly_parallel_blobs_do(&code_cl);

      } else {

        _rp->roots_do(worker_id, &roots_cl, &cld_cl, &code_cl);

      }

      AlwaysTrueClosure is_alive;

      _dedup_roots.oops_do(&is_alive, &roots_cl, worker_id);

    }

  }

};

class ShenandoahConcurrentTraversalCollectionTask : public AbstractGangTask {

private:

  ShenandoahTaskTerminator* _terminator;

  ShenandoahHeap* _heap;

public:

  ShenandoahConcurrentTraversalCollectionTask(ShenandoahTaskTerminator* terminator) :

    AbstractGangTask("Shenandoah Concurrent Traversal Collection"),

    _terminator(terminator),

    _heap(ShenandoahHeap::heap()) {}

  void work(uint worker_id) {

    ShenandoahConcurrentWorkerSession worker_session(worker_id);

    ShenandoahSuspendibleThreadSetJoiner stsj(ShenandoahSuspendibleWorkers);

    ShenandoahEvacOOMScope oom_evac_scope;

    ShenandoahTraversalGC* traversal_gc = _heap->traversal_gc();

    // Drain all outstanding work in queues.

    traversal_gc->main_loop(worker_id, _terminator, true);

  }

};

class ShenandoahFinalTraversalCollectionTask : public AbstractGangTask {

private:

  ShenandoahTaskTerminator* _terminator;

  ShenandoahHeap* _heap;

public:

    AbstractGangTask("Shenandoah Final Traversal Collection"),

    _rp(rp),

    _terminator(terminator),

    _heap(ShenandoahHeap::heap()) {}

  void work(uint worker_id) {

    ShenandoahParallelWorkerSession worker_session(worker_id);

    ShenandoahEvacOOMScope oom_evac_scope;

    ShenandoahTraversalGC* traversal_gc = _heap->traversal_gc();

    ShenandoahObjToScanQueueSet* queues = traversal_gc->task_queues();

    ShenandoahObjToScanQueue* q = queues->queue(worker_id);

    bool process_refs = _heap->process_references();

    bool unload_classes = _heap->unload_classes();

    ReferenceProcessor* rp = NULL;

    if (process_refs) {

      rp = _heap->ref_processor();

    }

    // Step 0: Drain outstanding SATB queues.

    // NOTE: we piggy-back draining of remaining thread SATB buffers on the final root scan below.

    ShenandoahTraversalSATBBufferClosure satb_cl(q);

    {

      // Process remaining finished SATB buffers.

      SATBMarkQueueSet& satb_mq_set = ShenandoahBarrierSet::satb_mark_queue_set();

      while (satb_mq_set.apply_closure_to_completed_buffer(&satb_cl));

      // Process remaining threads SATB buffers below.

    }

    // Step 1: Process GC roots.

    // For oops in code roots, they are marked, evacuated, enqueued for further traversal,

    // and the references to the oops are updated during init pause. New nmethods are handled

    // in similar way during nmethod-register process. Therefore, we don't need to rescan code

    // roots here.

    if (!_heap->is_degenerated_gc_in_progress()) {

      ShenandoahTraversalClosure roots_cl(q, rp);

      ShenandoahTraversalSATBThreadsClosure tc(&satb_cl);

      if (unload_classes) {

        ShenandoahRemarkCLDClosure remark_cld_cl(&roots_cl);

        _rp->strong_roots_do(worker_id, &roots_cl, &remark_cld_cl, NULL, &tc);

      } else {

        CLDToOopClosure cld_cl(&roots_cl, ClassLoaderData::_claim_strong);

        _rp->roots_do(worker_id, &roots_cl, &cld_cl, NULL, &tc);

      }

    } else {

      ShenandoahTraversalDegenClosure roots_cl(q, rp);

      ShenandoahTraversalSATBThreadsClosure tc(&satb_cl);

      if (unload_classes) {

        ShenandoahRemarkCLDClosure remark_cld_cl(&roots_cl);

        _rp->strong_roots_do(worker_id, &roots_cl, &remark_cld_cl, NULL, &tc);

      } else {

        CLDToOopClosure cld_cl(&roots_cl, ClassLoaderData::_claim_strong);

        _rp->roots_do(worker_id, &roots_cl, &cld_cl, NULL, &tc);

      }

    }

    {

      ShenandoahWorkerTimings *worker_times = _heap->phase_timings()->worker_times();

      ShenandoahWorkerTimingsTracker timer(worker_times, ShenandoahPhaseTimings::FinishQueues, worker_id);

      // Step 3: Finally drain all outstanding work in queues.

      traversal_gc->main_loop(worker_id, _terminator, false);

    }

  }

};

ShenandoahTraversalGC::ShenandoahTraversalGC(ShenandoahHeap* heap, size_t num_regions) :

  _heap(heap),

  _task_queues(new ShenandoahObjToScanQueueSet(heap->max_workers())),

  _traversal_set(ShenandoahHeapRegionSet()) {

  // Traversal does not support concurrent code root scanning

  FLAG_SET_DEFAULT(ShenandoahConcurrentScanCodeRoots, false);

  uint num_queues = heap->max_workers();

  for (uint i = 0; i < num_queues; ++i) {

    ShenandoahObjToScanQueue* task_queue = new ShenandoahObjToScanQueue();

    task_queue->initialize();

    _task_queues->register_queue(i, task_queue);

  }

}

ShenandoahTraversalGC::~ShenandoahTraversalGC() {

}

void ShenandoahTraversalGC::prepare_regions() {

  size_t num_regions = _heap->num_regions();

  ShenandoahMarkingContext* const ctx = _heap->marking_context();

  for (size_t i = 0; i < num_regions; i++) {

    ShenandoahHeapRegion* region = _heap->get_region(i);

    if (_heap->is_bitmap_slice_committed(region)) {

      if (_traversal_set.is_in(i)) {

        ctx->capture_top_at_mark_start(region);

        region->clear_live_data();

        assert(ctx->is_bitmap_clear_range(region->bottom(), region->end()), "bitmap for traversal regions must be cleared");

      } else {

        // Everything outside the traversal set is always considered live.

        ctx->reset_top_at_mark_start(region);

      }

    } else {

      // FreeSet may contain uncommitted empty regions, once they are recommitted,

      // their TAMS may have old values, so reset them here.

      ctx->reset_top_at_mark_start(region);

    }

  }

}

void ShenandoahTraversalGC::prepare() {

  _heap->collection_set()->clear();

  assert(_heap->collection_set()->count() == 0, "collection set not clear");

  {

    ShenandoahGCPhase phase(ShenandoahPhaseTimings::traversal_gc_make_parsable);

    _heap->make_parsable(true);

  }

  if (UseTLAB) {

    ShenandoahGCPhase phase(ShenandoahPhaseTimings::traversal_gc_resize_tlabs);

    _heap->resize_tlabs();

  }

  assert(_heap->marking_context()->is_bitmap_clear(), "need clean mark bitmap");

  assert(!_heap->marking_context()->is_complete(), "should not be complete");

  ShenandoahFreeSet* free_set = _heap->free_set();

  ShenandoahCollectionSet* collection_set = _heap->collection_set();

  // Find collection set

  _heap->heuristics()->choose_collection_set(collection_set);

  prepare_regions();

  // Rebuild free set

  free_set->rebuild();

  log_info(gc, ergo)("Collectable Garbage: " SIZE_FORMAT "M, " SIZE_FORMAT "M CSet, " SIZE_FORMAT " CSet regions",

                     collection_set->garbage() / M, collection_set->live_data() / M, collection_set->count());

}

void ShenandoahTraversalGC::init_traversal_collection() {

  assert(ShenandoahSafepoint::is_at_shenandoah_safepoint(), "STW traversal GC");

  if (ShenandoahVerify) {

    _heap->verifier()->verify_before_traversal();

  }

  if (VerifyBeforeGC) {

    Universe::verify();

  }

  {

    ShenandoahGCPhase phase_prepare(ShenandoahPhaseTimings::traversal_gc_prepare);

    ShenandoahHeapLocker lock(_heap->lock());

    prepare();

  }

  _heap->set_concurrent_traversal_in_progress(true);

  bool process_refs = _heap->process_references();

  if (process_refs) {

    ReferenceProcessor* rp = _heap->ref_processor();

    rp->enable_discovery(true /*verify_no_refs*/);

    rp->setup_policy(_heap->soft_ref_policy()->should_clear_all_soft_refs());

  }

  {

    ShenandoahGCPhase phase_work(ShenandoahPhaseTimings::init_traversal_gc_work);

    assert(_task_queues->is_empty(), "queues must be empty before traversal GC");

    TASKQUEUE_STATS_ONLY(_task_queues->reset_taskqueue_stats());

#if defined(COMPILER2) || INCLUDE_JVMCI

    DerivedPointerTable::clear();

#endif

    {

      uint nworkers = _heap->workers()->active_workers();

      task_queues()->reserve(nworkers);

      _heap->workers()->run_task(&traversal_task);

    }

#if defined(COMPILER2) || INCLUDE_JVMCI

    DerivedPointerTable::update_pointers();

#endif

  }

  if (ShenandoahPacing) {

    _heap->pacer()->setup_for_traversal();

  }

}

void ShenandoahTraversalGC::main_loop(uint w, ShenandoahTaskTerminator* t, bool sts_yield) {

  ShenandoahObjToScanQueue* q = task_queues()->queue(w);

  // Initialize live data.

  jushort* ld = _heap->get_liveness_cache(w);

  ReferenceProcessor* rp = NULL;

  if (_heap->process_references()) {

    rp = _heap->ref_processor();

  }

  {

    if (!_heap->is_degenerated_gc_in_progress()) {

      if (_heap->unload_classes()) {

        if (ShenandoahStringDedup::is_enabled()) {

          ShenandoahTraversalMetadataDedupClosure cl(q, rp);

          main_loop_work<ShenandoahTraversalMetadataDedupClosure>(&cl, ld, w, t, sts_yield);

        } else {

          ShenandoahTraversalMetadataClosure cl(q, rp);

          main_loop_work<ShenandoahTraversalMetadataClosure>(&cl, ld, w, t, sts_yield);

        }

      } else {

        if (ShenandoahStringDedup::is_enabled()) {

          ShenandoahTraversalDedupClosure cl(q, rp);

          main_loop_work<ShenandoahTraversalDedupClosure>(&cl, ld, w, t, sts_yield);

        } else {

          ShenandoahTraversalClosure cl(q, rp);

          main_loop_work<ShenandoahTraversalClosure>(&cl, ld, w, t, sts_yield);

        }

      }

    } else {

      if (_heap->unload_classes()) {

        if (ShenandoahStringDedup::is_enabled()) {

          ShenandoahTraversalMetadataDedupDegenClosure cl(q, rp);

          main_loop_work<ShenandoahTraversalMetadataDedupDegenClosure>(&cl, ld, w, t, sts_yield);

        } else {

          ShenandoahTraversalMetadataDegenClosure cl(q, rp);

          main_loop_work<ShenandoahTraversalMetadataDegenClosure>(&cl, ld, w, t, sts_yield);

        }

      } else {

        if (ShenandoahStringDedup::is_enabled()) {

          ShenandoahTraversalDedupDegenClosure cl(q, rp);

          main_loop_work<ShenandoahTraversalDedupDegenClosure>(&cl, ld, w, t, sts_yield);

        } else {

          ShenandoahTraversalDegenClosure cl(q, rp);

          main_loop_work<ShenandoahTraversalDegenClosure>(&cl, ld, w, t, sts_yield);

        }

      }

    }

  }

  _heap->flush_liveness_cache(w);

}

template <class T>

void ShenandoahTraversalGC::main_loop_work(T* cl, jushort* live_data, uint worker_id, ShenandoahTaskTerminator* terminator, bool sts_yield) {

  ShenandoahObjToScanQueueSet* queues = task_queues();

  ShenandoahObjToScanQueue* q = queues->queue(worker_id);

  ShenandoahConcurrentMark* conc_mark = _heap->concurrent_mark();

  uintx stride = ShenandoahMarkLoopStride;

  ShenandoahMarkTask task;

  // Process outstanding queues, if any.

  q = queues->claim_next();

  while (q != NULL) {

    if (_heap->check_cancelled_gc_and_yield(sts_yield)) {

      return;

    }

    for (uint i = 0; i < stride; i++) {

      if (q->pop(task)) {

        conc_mark->do_task<T>(q, cl, live_data, &task);

      } else {

        assert(q->is_empty(), "Must be empty");

        q = queues->claim_next();

        break;

      }

    }

  }

  if (check_and_handle_cancelled_gc(terminator, sts_yield)) return;

  // Normal loop.

  q = queues->queue(worker_id);

  ShenandoahTraversalSATBBufferClosure drain_satb(q);

  SATBMarkQueueSet& satb_mq_set = ShenandoahBarrierSet::satb_mark_queue_set();

  while (true) {

    if (check_and_handle_cancelled_gc(terminator, sts_yield)) return;