/* * Copyright (c) 2013, 2021, Red Hat, Inc. All rights reserved. * Copyright (C) 2022 THL A29 Limited, a Tencent company. All rights reserved. * Copyright Amazon.com Inc. 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/shenandoah/shenandoahAsserts.hpp" #include "gc/shenandoah/shenandoahCollectorPolicy.hpp" #include "gc/shenandoah/shenandoahConcurrentGC.hpp" #include "gc/shenandoah/shenandoahGenerationalControlThread.hpp" #include "gc/shenandoah/shenandoahDegeneratedGC.hpp" #include "gc/shenandoah/shenandoahFreeSet.hpp" #include "gc/shenandoah/shenandoahFullGC.hpp" #include "gc/shenandoah/shenandoahGeneration.hpp" #include "gc/shenandoah/shenandoahGenerationalHeap.hpp" #include "gc/shenandoah/shenandoahOldGC.hpp" #include "gc/shenandoah/shenandoahOldGeneration.hpp" #include "gc/shenandoah/shenandoahHeap.inline.hpp" #include "gc/shenandoah/shenandoahMonitoringSupport.hpp" #include "gc/shenandoah/shenandoahPacer.inline.hpp" #include "gc/shenandoah/shenandoahUtils.hpp" #include "gc/shenandoah/shenandoahYoungGeneration.hpp" #include "logging/log.hpp" #include "memory/metaspaceUtils.hpp" #include "memory/metaspaceStats.hpp" #include "runtime/atomic.hpp" #include "utilities/events.hpp" ShenandoahGenerationalControlThread::ShenandoahGenerationalControlThread() : _control_lock(Mutex::nosafepoint - 2, "ShenandoahGCRequest_lock", true), _requested_gc_cause(GCCause::_no_gc), _requested_generation(nullptr), _gc_mode(none), _degen_point(ShenandoahGC::_degenerated_unset), _heap(ShenandoahGenerationalHeap::heap()), _age_period(0) { shenandoah_assert_generational(); set_name("Shenandoah Control Thread"); create_and_start(); } void ShenandoahGenerationalControlThread::run_service() { const int64_t wait_ms = ShenandoahPacing ? ShenandoahControlIntervalMin : 0; ShenandoahGCRequest request; while (!should_terminate()) { // This control loop iteration has seen this much allocation. const size_t allocs_seen = reset_allocs_seen(); // Figure out if we have pending requests. check_for_request(request); if (request.cause == GCCause::_shenandoah_stop_vm) { break; } if (request.cause != GCCause::_no_gc) { run_gc_cycle(request); } else { // Report to pacer that we have seen this many words allocated if (ShenandoahPacing && (allocs_seen > 0)) { _heap->pacer()->report_alloc(allocs_seen); } } // If the cycle was cancelled, continue the next iteration to deal with it. Otherwise, // if there was no other cycle requested, cleanup and wait for the next request. if (!_heap->cancelled_gc()) { MonitorLocker ml(&_control_lock, Mutex::_no_safepoint_check_flag); if (_requested_gc_cause == GCCause::_no_gc) { set_gc_mode(ml, none); ml.wait(wait_ms); } } } // In case any threads are waiting for a cycle to happen, notify them so they observe the shutdown. notify_gc_waiters(); notify_alloc_failure_waiters(); set_gc_mode(stopped); } void ShenandoahGenerationalControlThread::stop_service() { log_debug(gc, thread)("Stopping control thread"); MonitorLocker ml(&_control_lock, Mutex::_no_safepoint_check_flag); _heap->cancel_gc(GCCause::_shenandoah_stop_vm); _requested_gc_cause = GCCause::_shenandoah_stop_vm; notify_cancellation(ml, GCCause::_shenandoah_stop_vm); // We can't wait here because it may interfere with the active cycle's ability // to reach a safepoint (this runs on a java thread). } void ShenandoahGenerationalControlThread::check_for_request(ShenandoahGCRequest& request) { // Hold the lock while we read request cause and generation MonitorLocker ml(&_control_lock, Mutex::_no_safepoint_check_flag); if (_heap->cancelled_gc()) { // The previous request was cancelled. Either it was cancelled for an allocation // failure (degenerated cycle), or old marking was cancelled to run a young collection. // In either case, the correct generation for the next cycle can be determined by // the cancellation cause. request.cause = _heap->cancelled_cause(); if (request.cause == GCCause::_shenandoah_concurrent_gc) { request.generation = _heap->young_generation(); _heap->clear_cancelled_gc(false); } } else { request.cause = _requested_gc_cause; request.generation = _requested_generation; // Only clear these if we made a request from them. In the case of a cancelled gc, // we do not want to inadvertently lose this pending request. _requested_gc_cause = GCCause::_no_gc; _requested_generation = nullptr; } if (request.cause == GCCause::_no_gc || request.cause == GCCause::_shenandoah_stop_vm) { return; } GCMode mode; if (ShenandoahCollectorPolicy::is_allocation_failure(request.cause)) { mode = prepare_for_allocation_failure_gc(request); } else if (ShenandoahCollectorPolicy::is_explicit_gc(request.cause)) { mode = prepare_for_explicit_gc(request); } else { mode = prepare_for_concurrent_gc(request); } set_gc_mode(ml, mode); } ShenandoahGenerationalControlThread::GCMode ShenandoahGenerationalControlThread::prepare_for_allocation_failure_gc(ShenandoahGCRequest &request) { if (_degen_point == ShenandoahGC::_degenerated_unset) { _degen_point = ShenandoahGC::_degenerated_outside_cycle; request.generation = _heap->young_generation(); } else if (request.generation->is_old()) { // This means we degenerated during the young bootstrap for the old generation // cycle. The following degenerated cycle should therefore also be young. request.generation = _heap->young_generation(); } ShenandoahHeuristics* heuristics = request.generation->heuristics(); bool old_gen_evacuation_failed = _heap->old_generation()->clear_failed_evacuation(); heuristics->log_trigger("Handle Allocation Failure"); // Do not bother with degenerated cycle if old generation evacuation failed or if humongous allocation failed if (ShenandoahDegeneratedGC && heuristics->should_degenerate_cycle() && !old_gen_evacuation_failed && request.cause != GCCause::_shenandoah_humongous_allocation_failure) { heuristics->record_allocation_failure_gc(); _heap->shenandoah_policy()->record_alloc_failure_to_degenerated(_degen_point); return stw_degenerated; } else { heuristics->record_allocation_failure_gc(); _heap->shenandoah_policy()->record_alloc_failure_to_full(); request.generation = _heap->global_generation(); return stw_full; } } ShenandoahGenerationalControlThread::GCMode ShenandoahGenerationalControlThread::prepare_for_explicit_gc(ShenandoahGCRequest &request) const { ShenandoahHeuristics* global_heuristics = _heap->global_generation()->heuristics(); request.generation = _heap->global_generation(); global_heuristics->log_trigger("GC request (%s)", GCCause::to_string(request.cause)); global_heuristics->record_requested_gc(); if (ShenandoahCollectorPolicy::should_run_full_gc(request.cause)) { return stw_full;; } else { // Unload and clean up everything. Note that this is an _explicit_ request and so does not use // the same `should_unload_classes` call as the regulator's concurrent gc request. _heap->set_unload_classes(global_heuristics->can_unload_classes()); return concurrent_normal; } } ShenandoahGenerationalControlThread::GCMode ShenandoahGenerationalControlThread::prepare_for_concurrent_gc(const ShenandoahGCRequest &request) const { assert(!(request.generation->is_old() && _heap->old_generation()->is_doing_mixed_evacuations()), "Old heuristic should not request cycles while it waits for mixed evacuations"); if (request.generation->is_global()) { ShenandoahHeuristics* global_heuristics = _heap->global_generation()->heuristics(); _heap->set_unload_classes(global_heuristics->should_unload_classes()); } else { _heap->set_unload_classes(false); } // preemption was requested or this is a regular cycle return request.generation->is_old() ? servicing_old : concurrent_normal; } void ShenandoahGenerationalControlThread::maybe_set_aging_cycle() { if (_age_period-- == 0) { _heap->set_aging_cycle(true); _age_period = ShenandoahAgingCyclePeriod - 1; } else { _heap->set_aging_cycle(false); } } void ShenandoahGenerationalControlThread::run_gc_cycle(const ShenandoahGCRequest& request) { log_debug(gc, thread)("Starting GC (%s): %s, %s", gc_mode_name(gc_mode()), GCCause::to_string(request.cause), request.generation->name()); assert(gc_mode() != none, "GC mode cannot be none here"); // Blow away all soft references on this cycle, if handling allocation failure, // either implicit or explicit GC request, or we are requested to do so unconditionally. if (request.generation->is_global() && (ShenandoahCollectorPolicy::is_allocation_failure(request.cause) || ShenandoahCollectorPolicy::is_explicit_gc(request.cause) || ShenandoahAlwaysClearSoftRefs)) { _heap->soft_ref_policy()->set_should_clear_all_soft_refs(true); } // GC is starting, bump the internal ID update_gc_id(); _heap->reset_bytes_allocated_since_gc_start(); MetaspaceCombinedStats meta_sizes = MetaspaceUtils::get_combined_statistics(); // If GC was requested, we are sampling the counters even without actual triggers // from allocation machinery. This captures GC phases more accurately. _heap->set_forced_counters_update(true); // If GC was requested, we better dump freeset data for performance debugging _heap->free_set()->log_status_under_lock(); { // Cannot uncommit bitmap slices during concurrent reset ShenandoahNoUncommitMark forbid_region_uncommit(_heap); switch (gc_mode()) { case concurrent_normal: { service_concurrent_normal_cycle(request); break; } case stw_degenerated: { service_stw_degenerated_cycle(request); break; } case stw_full: { service_stw_full_cycle(request.cause); break; } case servicing_old: { assert(request.generation->is_old(), "Expected old generation here"); GCIdMark gc_id_mark; service_concurrent_old_cycle(request); break; } default: ShouldNotReachHere(); } } // If this was the requested GC cycle, notify waiters about it if (ShenandoahCollectorPolicy::is_explicit_gc(request.cause)) { notify_gc_waiters(); } // If this was an allocation failure GC cycle, notify waiters about it if (ShenandoahCollectorPolicy::is_allocation_failure(request.cause)) { notify_alloc_failure_waiters(); } // Report current free set state at the end of cycle, whether // it is a normal completion, or the abort. _heap->free_set()->log_status_under_lock(); // Notify Universe about new heap usage. This has implications for // global soft refs policy, and we better report it every time heap // usage goes down. _heap->update_capacity_and_used_at_gc(); // Signal that we have completed a visit to all live objects. _heap->record_whole_heap_examined_timestamp(); // Disable forced counters update, and update counters one more time // to capture the state at the end of GC session. _heap->handle_force_counters_update(); _heap->set_forced_counters_update(false); // Retract forceful part of soft refs policy _heap->soft_ref_policy()->set_should_clear_all_soft_refs(false); // Clear metaspace oom flag, if current cycle unloaded classes if (_heap->unload_classes()) { _heap->global_generation()->heuristics()->clear_metaspace_oom(); } process_phase_timings(); // Print Metaspace change following GC (if logging is enabled). MetaspaceUtils::print_metaspace_change(meta_sizes); // GC is over, we are at idle now if (ShenandoahPacing) { _heap->pacer()->setup_for_idle(); } // Check if we have seen a new target for soft max heap size or if a gc was requested. // Either of these conditions will attempt to uncommit regions. if (ShenandoahUncommit) { if (_heap->check_soft_max_changed()) { _heap->notify_soft_max_changed(); } else if (ShenandoahCollectorPolicy::is_explicit_gc(request.cause)) { _heap->notify_explicit_gc_requested(); } } log_debug(gc, thread)("Completed GC (%s): %s, %s, cancelled: %s", gc_mode_name(gc_mode()), GCCause::to_string(request.cause), request.generation->name(), GCCause::to_string(_heap->cancelled_cause())); } void ShenandoahGenerationalControlThread::process_phase_timings() const { // Commit worker statistics to cycle data _heap->phase_timings()->flush_par_workers_to_cycle(); if (ShenandoahPacing) { _heap->pacer()->flush_stats_to_cycle(); } ShenandoahEvacuationTracker* evac_tracker = _heap->evac_tracker(); ShenandoahCycleStats evac_stats = evac_tracker->flush_cycle_to_global(); // Print GC stats for current cycle { LogTarget(Info, gc, stats) lt; if (lt.is_enabled()) { ResourceMark rm; LogStream ls(lt); _heap->phase_timings()->print_cycle_on(&ls); evac_tracker->print_evacuations_on(&ls, &evac_stats.workers, &evac_stats.mutators); if (ShenandoahPacing) { _heap->pacer()->print_cycle_on(&ls); } } } // Commit statistics to globals _heap->phase_timings()->flush_cycle_to_global(); } // Young and old concurrent cycles are initiated by the regulator. Implicit // and explicit GC requests are handled by the controller thread and always // run a global cycle (which is concurrent by default, but may be overridden // by command line options). Old cycles always degenerate to a global cycle. // Young cycles are degenerated to complete the young cycle. Young // and old degen may upgrade to Full GC. Full GC may also be // triggered directly by a System.gc() invocation. // // // +-----+ Idle +-----+-----------+---------------------+ // | + | | | // | | | | | // | | v | | // | | Bootstrap Old +-- | ------------+ | // | | + | | | // | | | | | | // | v v v v | // | Resume Old <----------+ Young +--> Young Degen | // | + + ^ + + | // v | | | | | | // Global <-+ | +----------------------------+ | | // + | | | // | v v | // +---> Global Degen +--------------------> Full <----+ // void ShenandoahGenerationalControlThread::service_concurrent_normal_cycle(const ShenandoahGCRequest& request) { GCIdMark gc_id_mark; log_info(gc, ergo)("Start GC cycle (%s)", request.generation->name()); if (request.generation->is_old()) { service_concurrent_old_cycle(request); } else { service_concurrent_cycle(request.generation, request.cause, false); } } void ShenandoahGenerationalControlThread::service_concurrent_old_cycle(const ShenandoahGCRequest& request) { ShenandoahOldGeneration* old_generation = _heap->old_generation(); ShenandoahYoungGeneration* young_generation = _heap->young_generation(); ShenandoahOldGeneration::State original_state = old_generation->state(); TraceCollectorStats tcs(_heap->monitoring_support()->concurrent_collection_counters()); switch (original_state) { case ShenandoahOldGeneration::FILLING: { ShenandoahGCSession session(request.cause, old_generation); assert(gc_mode() == servicing_old, "Filling should be servicing old"); _allow_old_preemption.set(); old_generation->entry_coalesce_and_fill(); _allow_old_preemption.unset(); // Before bootstrapping begins, we must acknowledge any cancellation request. // If the gc has not been cancelled, this does nothing. If it has been cancelled, // this will clear the cancellation request and exit before starting the bootstrap // phase. This will allow the young GC cycle to proceed normally. If we do not // acknowledge the cancellation request, the subsequent young cycle will observe // the request and essentially cancel itself. if (check_cancellation_or_degen(ShenandoahGC::_degenerated_outside_cycle)) { log_info(gc, thread)("Preparation for old generation cycle was cancelled"); return; } // Coalescing threads completed and nothing was cancelled. it is safe to transition from this state. old_generation->transition_to(ShenandoahOldGeneration::WAITING_FOR_BOOTSTRAP); return; } case ShenandoahOldGeneration::WAITING_FOR_BOOTSTRAP: old_generation->transition_to(ShenandoahOldGeneration::BOOTSTRAPPING); case ShenandoahOldGeneration::BOOTSTRAPPING: { // Configure the young generation's concurrent mark to put objects in // old regions into the concurrent mark queues associated with the old // generation. The young cycle will run as normal except that rather than // ignore old references it will mark and enqueue them in the old concurrent // task queues but it will not traverse them. set_gc_mode(bootstrapping_old); young_generation->set_old_gen_task_queues(old_generation->task_queues()); service_concurrent_cycle(young_generation, request.cause, true); process_phase_timings(); if (_heap->cancelled_gc()) { // Young generation bootstrap cycle has failed. Concurrent mark for old generation // is going to resume after degenerated bootstrap cycle completes. log_info(gc)("Bootstrap cycle for old generation was cancelled"); return; } assert(_degen_point == ShenandoahGC::_degenerated_unset, "Degen point should not be set if gc wasn't cancelled"); // From here we will 'resume' the old concurrent mark. This will skip reset // and init mark for the concurrent mark. All of that work will have been // done by the bootstrapping young cycle. set_gc_mode(servicing_old); old_generation->transition_to(ShenandoahOldGeneration::MARKING); } case ShenandoahOldGeneration::MARKING: { ShenandoahGCSession session(request.cause, old_generation); bool marking_complete = resume_concurrent_old_cycle(old_generation, request.cause); if (marking_complete) { assert(old_generation->state() != ShenandoahOldGeneration::MARKING, "Should not still be marking"); if (original_state == ShenandoahOldGeneration::MARKING) { _heap->mmu_tracker()->record_old_marking_increment(true); _heap->log_heap_status("At end of Concurrent Old Marking finishing increment"); } } else if (original_state == ShenandoahOldGeneration::MARKING) { _heap->mmu_tracker()->record_old_marking_increment(false); _heap->log_heap_status("At end of Concurrent Old Marking increment"); } break; } default: fatal("Unexpected state for old GC: %s", ShenandoahOldGeneration::state_name(old_generation->state())); } } bool ShenandoahGenerationalControlThread::resume_concurrent_old_cycle(ShenandoahOldGeneration* generation, GCCause::Cause cause) { assert(_heap->is_concurrent_old_mark_in_progress(), "Old mark should be in progress"); log_debug(gc)("Resuming old generation with " UINT32_FORMAT " marking tasks queued", generation->task_queues()->tasks()); // We can only tolerate being cancelled during concurrent marking or during preparation for mixed // evacuation. This flag here (passed by reference) is used to control precisely where the regulator // is allowed to cancel a GC. ShenandoahOldGC gc(generation, _allow_old_preemption); if (gc.collect(cause)) { _heap->notify_gc_progress(); generation->record_success_concurrent(false); } if (_heap->cancelled_gc()) { // It's possible the gc cycle was cancelled after the last time // the collection checked for cancellation. In which case, the // old gc cycle is still completed, and we have to deal with this // cancellation. We set the degeneration point to be outside // the cycle because if this is an allocation failure, that is // what must be done (there is no degenerated old cycle). If the // cancellation was due to a heuristic wanting to start a young // cycle, then we are not actually going to a degenerated cycle, // so the degenerated point doesn't matter here. check_cancellation_or_degen(ShenandoahGC::_degenerated_outside_cycle); if (cause == GCCause::_shenandoah_concurrent_gc) { _heap->shenandoah_policy()->record_interrupted_old(); } return false; } return true; } // Normal cycle goes via all concurrent phases. If allocation failure (af) happens during // any of the concurrent phases, it first degrades to Degenerated GC and completes GC there. // If second allocation failure happens during Degenerated GC cycle (for example, when GC // tries to evac something and no memory is available), cycle degrades to Full GC. // // There are also a shortcut through the normal cycle: immediate garbage shortcut, when // heuristics says there are no regions to compact, and all the collection comes from immediately // reclaimable regions. // // ................................................................................................ // // (immediate garbage shortcut) Concurrent GC // /-------------------------------------------\ // | | // | | // | | // | v // [START] ----> Conc Mark ----o----> Conc Evac --o--> Conc Update-Refs ---o----> [END] // | | | ^ // | (af) | (af) | (af) | // ..................|....................|.................|..............|....................... // | | | | // | | | | Degenerated GC // v v v | // STW Mark ----------> STW Evac ----> STW Update-Refs ----->o // | | | ^ // | (af) | (af) | (af) | // ..................|....................|.................|..............|....................... // | | | | // | v | | Full GC // \------------------->o<----------------/ | // | | // v | // Full GC --------------------------/ // void ShenandoahGenerationalControlThread::service_concurrent_cycle(ShenandoahGeneration* generation, GCCause::Cause cause, bool do_old_gc_bootstrap) { // At this point: // if (generation == YOUNG), this is a normal young cycle or a bootstrap cycle // if (generation == GLOBAL), this is a GLOBAL cycle // In either case, we want to age old objects if this is an aging cycle maybe_set_aging_cycle(); ShenandoahGCSession session(cause, generation); TraceCollectorStats tcs(_heap->monitoring_support()->concurrent_collection_counters()); assert(!generation->is_old(), "Old GC takes a different control path"); ShenandoahConcurrentGC gc(generation, do_old_gc_bootstrap); if (gc.collect(cause)) { // Cycle is complete _heap->notify_gc_progress(); generation->record_success_concurrent(gc.abbreviated()); } else { assert(_heap->cancelled_gc(), "Must have been cancelled"); check_cancellation_or_degen(gc.degen_point()); } const char* msg; ShenandoahMmuTracker* mmu_tracker = _heap->mmu_tracker(); if (generation->is_young()) { if (_heap->cancelled_gc()) { msg = (do_old_gc_bootstrap) ? "At end of Interrupted Concurrent Bootstrap GC" : "At end of Interrupted Concurrent Young GC"; } else { // We only record GC results if GC was successful msg = (do_old_gc_bootstrap) ? "At end of Concurrent Bootstrap GC" : "At end of Concurrent Young GC"; if (_heap->collection_set()->has_old_regions()) { mmu_tracker->record_mixed(get_gc_id()); } else if (do_old_gc_bootstrap) { mmu_tracker->record_bootstrap(get_gc_id()); } else { mmu_tracker->record_young(get_gc_id()); } } } else { assert(generation->is_global(), "If not young, must be GLOBAL"); assert(!do_old_gc_bootstrap, "Do not bootstrap with GLOBAL GC"); if (_heap->cancelled_gc()) { msg = "At end of Interrupted Concurrent GLOBAL GC"; } else { // We only record GC results if GC was successful msg = "At end of Concurrent Global GC"; mmu_tracker->record_global(get_gc_id()); } } _heap->log_heap_status(msg); } bool ShenandoahGenerationalControlThread::check_cancellation_or_degen(ShenandoahGC::ShenandoahDegenPoint point) { if (!_heap->cancelled_gc()) { return false; } if (_heap->cancelled_cause() == GCCause::_shenandoah_stop_vm || _heap->cancelled_cause() == GCCause::_shenandoah_concurrent_gc) { log_debug(gc, thread)("Cancellation detected, reason: %s", GCCause::to_string(_heap->cancelled_cause())); return true; } if (ShenandoahCollectorPolicy::is_allocation_failure(_heap->cancelled_cause())) { assert(_degen_point == ShenandoahGC::_degenerated_unset, "Should not be set yet: %s", ShenandoahGC::degen_point_to_string(_degen_point)); _degen_point = point; log_debug(gc, thread)("Cancellation detected:, reason: %s, degen point: %s", GCCause::to_string(_heap->cancelled_cause()), ShenandoahGC::degen_point_to_string(_degen_point)); return true; } fatal("Cancel GC either for alloc failure GC, or gracefully exiting, or to pause old generation marking"); return false; } void ShenandoahGenerationalControlThread::service_stw_full_cycle(GCCause::Cause cause) { GCIdMark gc_id_mark; ShenandoahGCSession session(cause, _heap->global_generation()); maybe_set_aging_cycle(); ShenandoahFullGC gc; gc.collect(cause); _degen_point = ShenandoahGC::_degenerated_unset; } void ShenandoahGenerationalControlThread::service_stw_degenerated_cycle(const ShenandoahGCRequest& request) { assert(_degen_point != ShenandoahGC::_degenerated_unset, "Degenerated point should be set"); GCIdMark gc_id_mark; ShenandoahGCSession session(request.cause, request.generation); ShenandoahDegenGC gc(_degen_point, request.generation); gc.collect(request.cause); _degen_point = ShenandoahGC::_degenerated_unset; assert(_heap->young_generation()->task_queues()->is_empty(), "Unexpected young generation marking tasks"); if (request.generation->is_global()) { assert(_heap->old_generation()->task_queues()->is_empty(), "Unexpected old generation marking tasks"); assert(_heap->global_generation()->task_queues()->is_empty(), "Unexpected global generation marking tasks"); } else { assert(request.generation->is_young(), "Expected degenerated young cycle, if not global."); ShenandoahOldGeneration* old = _heap->old_generation(); if (old->is_bootstrapping()) { old->transition_to(ShenandoahOldGeneration::MARKING); } } } void ShenandoahGenerationalControlThread::request_gc(GCCause::Cause cause) { if (ShenandoahCollectorPolicy::is_allocation_failure(cause)) { // GC should already be cancelled. Here we are just notifying the control thread to // wake up and handle the cancellation request, so we don't need to set _requested_gc_cause. notify_cancellation(cause); } else if (ShenandoahCollectorPolicy::should_handle_requested_gc(cause)) { handle_requested_gc(cause); } } bool ShenandoahGenerationalControlThread::request_concurrent_gc(ShenandoahGeneration* generation) { if (_heap->cancelled_gc()) { // Ignore subsequent requests from the heuristics log_debug(gc, thread)("Reject request for concurrent gc: gc_requested: %s, gc_cancelled: %s", GCCause::to_string(_requested_gc_cause), BOOL_TO_STR(_heap->cancelled_gc())); return false; } MonitorLocker ml(&_control_lock, Mutex::_no_safepoint_check_flag); if (gc_mode() == servicing_old) { if (!preempt_old_marking(generation)) { log_debug(gc, thread)("Cannot start young, old collection is not preemptible"); return false; } // Cancel the old GC and wait for the control thread to start servicing the new request. log_info(gc)("Preempting old generation mark to allow %s GC", generation->name()); while (gc_mode() == servicing_old) { ShenandoahHeap::heap()->cancel_gc(GCCause::_shenandoah_concurrent_gc); notify_cancellation(ml, GCCause::_shenandoah_concurrent_gc); ml.wait(); } return true; } if (gc_mode() == none) { while (gc_mode() == none) { if (_requested_gc_cause != GCCause::_no_gc) { log_debug(gc, thread)("Reject request for concurrent gc because another gc is pending: %s", GCCause::to_string(_requested_gc_cause)); return false; } notify_control_thread(ml, GCCause::_shenandoah_concurrent_gc, generation); ml.wait(); } return true; } log_debug(gc, thread)("Reject request for concurrent gc: mode: %s, allow_old_preemption: %s", gc_mode_name(gc_mode()), BOOL_TO_STR(_allow_old_preemption.is_set())); return false; } void ShenandoahGenerationalControlThread::notify_control_thread(GCCause::Cause cause, ShenandoahGeneration* generation) { MonitorLocker ml(&_control_lock, Mutex::_no_safepoint_check_flag); notify_control_thread(ml, cause, generation); } void ShenandoahGenerationalControlThread::notify_control_thread(MonitorLocker& ml, GCCause::Cause cause, ShenandoahGeneration* generation) { assert(_control_lock.is_locked(), "Request lock must be held here"); log_debug(gc, thread)("Notify control (%s): %s, %s", gc_mode_name(gc_mode()), GCCause::to_string(cause), generation->name()); _requested_gc_cause = cause; _requested_generation = generation; ml.notify(); } void ShenandoahGenerationalControlThread::notify_cancellation(GCCause::Cause cause) { MonitorLocker ml(&_control_lock, Mutex::_no_safepoint_check_flag); notify_cancellation(ml, cause); } void ShenandoahGenerationalControlThread::notify_cancellation(MonitorLocker& ml, GCCause::Cause cause) { assert(_heap->cancelled_gc(), "GC should already be cancelled"); log_debug(gc,thread)("Notify control (%s): %s", gc_mode_name(gc_mode()), GCCause::to_string(cause)); ml.notify(); } bool ShenandoahGenerationalControlThread::preempt_old_marking(ShenandoahGeneration* generation) { return generation->is_young() && _allow_old_preemption.try_unset(); } void ShenandoahGenerationalControlThread::handle_requested_gc(GCCause::Cause cause) { // For normal requested GCs (System.gc) we want to block the caller. However, // for whitebox requested GC, we want to initiate the GC and return immediately. // The whitebox caller thread will arrange for itself to wait until the GC notifies // it that has reached the requested breakpoint (phase in the GC). if (cause == GCCause::_wb_breakpoint) { notify_control_thread(cause, ShenandoahHeap::heap()->global_generation()); return; } // Make sure we have at least one complete GC cycle before unblocking // from the explicit GC request. // // This is especially important for weak references cleanup and/or native // resources (e.g. DirectByteBuffers) machinery: when explicit GC request // comes very late in the already running cycle, it would miss lots of new // opportunities for cleanup that were made available before the caller // requested the GC. MonitorLocker ml(&_gc_waiters_lock); size_t current_gc_id = get_gc_id(); const size_t required_gc_id = current_gc_id + 1; while (current_gc_id < required_gc_id && !should_terminate()) { // Make requests to run a global cycle until at least one is completed notify_control_thread(cause, ShenandoahHeap::heap()->global_generation()); ml.wait(); current_gc_id = get_gc_id(); } } void ShenandoahGenerationalControlThread::notify_gc_waiters() { MonitorLocker ml(&_gc_waiters_lock); ml.notify_all(); } const char* ShenandoahGenerationalControlThread::gc_mode_name(GCMode mode) { switch (mode) { case none: return "idle"; case concurrent_normal: return "normal"; case stw_degenerated: return "degenerated"; case stw_full: return "full"; case servicing_old: return "old"; case bootstrapping_old: return "bootstrap"; case stopped: return "stopped"; default: return "unknown"; } } void ShenandoahGenerationalControlThread::set_gc_mode(GCMode new_mode) { MonitorLocker ml(&_control_lock, Mutex::_no_safepoint_check_flag); set_gc_mode(ml, new_mode); } void ShenandoahGenerationalControlThread::set_gc_mode(MonitorLocker& ml, GCMode new_mode) { if (_gc_mode != new_mode) { log_debug(gc, thread)("Transition from: %s to: %s", gc_mode_name(_gc_mode), gc_mode_name(new_mode)); EventMark event("Control thread transition from: %s, to %s", gc_mode_name(_gc_mode), gc_mode_name(new_mode)); _gc_mode = new_mode; ml.notify_all(); } }