blob: 18c7038fb1ea1749f7d8f97eda96e5888a9cb2d6 [file] [log] [blame]
// Copyright 2016 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "base/task/task_scheduler/scheduler_worker_pool_impl.h"
#include <memory>
#include <unordered_set>
#include <vector>
#include "base/atomicops.h"
#include "base/barrier_closure.h"
#include "base/bind.h"
#include "base/bind_helpers.h"
#include "base/callback.h"
#include "base/macros.h"
#include "base/memory/ptr_util.h"
#include "base/memory/ref_counted.h"
#include "base/metrics/histogram.h"
#include "base/metrics/histogram_samples.h"
#include "base/metrics/statistics_recorder.h"
#include "base/synchronization/atomic_flag.h"
#include "base/synchronization/condition_variable.h"
#include "base/synchronization/lock.h"
#include "base/synchronization/waitable_event.h"
#include "base/task/task_scheduler/delayed_task_manager.h"
#include "base/task/task_scheduler/scheduler_worker_pool_params.h"
#include "base/task/task_scheduler/sequence.h"
#include "base/task/task_scheduler/sequence_sort_key.h"
#include "base/task/task_scheduler/task_tracker.h"
#include "base/task/task_scheduler/test_task_factory.h"
#include "base/task/task_scheduler/test_utils.h"
#include "base/task_runner.h"
#include "base/test/bind_test_util.h"
#include "base/test/gtest_util.h"
#include "base/test/metrics/histogram_tester.h"
#include "base/test/test_simple_task_runner.h"
#include "base/test/test_timeouts.h"
#include "base/threading/platform_thread.h"
#include "base/threading/scoped_blocking_call.h"
#include "base/threading/simple_thread.h"
#include "base/threading/thread.h"
#include "base/threading/thread_checker_impl.h"
#include "base/threading/thread_local_storage.h"
#include "base/threading/thread_restrictions.h"
#include "base/time/time.h"
#include "base/timer/timer.h"
#include "build/build_config.h"
#include "starboard/configuration_constants.h"
#include "testing/gtest/include/gtest/gtest.h"
#if defined(OS_WIN)
#include "base/win/com_init_util.h"
#include "starboard/types.h"
#endif // defined(OS_WIN)
namespace base {
namespace internal {
namespace {
constexpr size_t kMaxTasks = 4;
constexpr size_t kNumThreadsPostingTasks = 4;
constexpr size_t kNumTasksPostedPerThread = 150;
// This can't be lower because Windows' WaitableEvent wakes up too early when a
// small timeout is used. This results in many spurious wake ups before a worker
// is allowed to cleanup.
constexpr TimeDelta kReclaimTimeForCleanupTests =
TimeDelta::FromMilliseconds(500);
// Waits on |event| in a scope where the blocking observer is null, to avoid
// affecting the max tasks.
void WaitWithoutBlockingObserver(WaitableEvent* event) {
internal::ScopedClearBlockingObserverForTesting clear_blocking_observer;
ScopedAllowBaseSyncPrimitivesForTesting allow_base_sync_primitives;
event->Wait();
}
class TaskSchedulerWorkerPoolImplTestBase {
protected:
TaskSchedulerWorkerPoolImplTestBase()
: statistics_recorder_(StatisticsRecorder::CreateTemporaryForTesting()),
service_thread_("TaskSchedulerServiceThread"){};
void CommonSetUp(TimeDelta suggested_reclaim_time = TimeDelta::Max()) {
CreateAndStartWorkerPool(suggested_reclaim_time, kMaxTasks);
}
void CommonTearDown() {
service_thread_.Stop();
task_tracker_.FlushForTesting();
if (worker_pool_)
worker_pool_->JoinForTesting();
}
void CreateWorkerPool() {
ASSERT_FALSE(worker_pool_);
service_thread_.Start();
delayed_task_manager_.Start(service_thread_.task_runner());
worker_pool_ = std::make_unique<SchedulerWorkerPoolImpl>(
"TestWorkerPool", "A", ThreadPriority::NORMAL,
task_tracker_.GetTrackedRef(), &delayed_task_manager_);
ASSERT_TRUE(worker_pool_);
}
virtual void StartWorkerPool(TimeDelta suggested_reclaim_time,
size_t max_tasks) {
ASSERT_TRUE(worker_pool_);
worker_pool_->Start(
SchedulerWorkerPoolParams(max_tasks, suggested_reclaim_time), max_tasks,
service_thread_.task_runner(), nullptr,
SchedulerWorkerPoolImpl::WorkerEnvironment::NONE);
}
void CreateAndStartWorkerPool(TimeDelta suggested_reclaim_time,
size_t max_tasks) {
CreateWorkerPool();
StartWorkerPool(suggested_reclaim_time, max_tasks);
}
std::unique_ptr<StatisticsRecorder> statistics_recorder_;
Thread service_thread_;
TaskTracker task_tracker_ = {"Test"};
std::unique_ptr<SchedulerWorkerPoolImpl> worker_pool_;
private:
DelayedTaskManager delayed_task_manager_;
DISALLOW_COPY_AND_ASSIGN(TaskSchedulerWorkerPoolImplTestBase);
};
class TaskSchedulerWorkerPoolImplTest
: public TaskSchedulerWorkerPoolImplTestBase,
public testing::Test {
protected:
TaskSchedulerWorkerPoolImplTest() = default;
void SetUp() override { TaskSchedulerWorkerPoolImplTestBase::CommonSetUp(); }
void TearDown() override {
TaskSchedulerWorkerPoolImplTestBase::CommonTearDown();
}
private:
DISALLOW_COPY_AND_ASSIGN(TaskSchedulerWorkerPoolImplTest);
};
class TaskSchedulerWorkerPoolImplTestParam
: public TaskSchedulerWorkerPoolImplTestBase,
public testing::TestWithParam<test::ExecutionMode> {
protected:
TaskSchedulerWorkerPoolImplTestParam() = default;
void SetUp() override { TaskSchedulerWorkerPoolImplTestBase::CommonSetUp(); }
void TearDown() override {
TaskSchedulerWorkerPoolImplTestBase::CommonTearDown();
}
private:
DISALLOW_COPY_AND_ASSIGN(TaskSchedulerWorkerPoolImplTestParam);
};
using PostNestedTask = test::TestTaskFactory::PostNestedTask;
class ThreadPostingTasksWaitIdle : public SimpleThread {
public:
// Constructs a thread that posts tasks to |worker_pool| through an
// |execution_mode| task runner. The thread waits until all workers in
// |worker_pool| are idle before posting a new task.
ThreadPostingTasksWaitIdle(SchedulerWorkerPoolImpl* worker_pool,
test::ExecutionMode execution_mode)
: SimpleThread("ThreadPostingTasksWaitIdle"),
worker_pool_(worker_pool),
#ifdef STARBOARD
task_runner_(nullptr),
#endif
factory_(CreateTaskRunnerWithExecutionMode(worker_pool, execution_mode),
execution_mode) {
DCHECK(worker_pool_);
}
const test::TestTaskFactory* factory() const { return &factory_; }
private:
void Run() override {
EXPECT_FALSE(factory_.task_runner()->RunsTasksInCurrentSequence());
for (size_t i = 0; i < kNumTasksPostedPerThread; ++i) {
worker_pool_->WaitForAllWorkersIdleForTesting();
EXPECT_TRUE(factory_.PostTask(PostNestedTask::NO, Closure()));
}
}
SchedulerWorkerPoolImpl* const worker_pool_;
const scoped_refptr<TaskRunner> task_runner_;
test::TestTaskFactory factory_;
DISALLOW_COPY_AND_ASSIGN(ThreadPostingTasksWaitIdle);
};
} // namespace
TEST_P(TaskSchedulerWorkerPoolImplTestParam, PostTasksWaitAllWorkersIdle) {
// Create threads to post tasks. To verify that workers can sleep and be woken
// up when new tasks are posted, wait for all workers to become idle before
// posting a new task.
std::vector<std::unique_ptr<ThreadPostingTasksWaitIdle>>
threads_posting_tasks;
for (size_t i = 0; i < kNumThreadsPostingTasks; ++i) {
threads_posting_tasks.push_back(
std::make_unique<ThreadPostingTasksWaitIdle>(worker_pool_.get(),
GetParam()));
threads_posting_tasks.back()->Start();
}
// Wait for all tasks to run.
for (const auto& thread_posting_tasks : threads_posting_tasks) {
thread_posting_tasks->Join();
thread_posting_tasks->factory()->WaitForAllTasksToRun();
}
// Wait until all workers are idle to be sure that no task accesses its
// TestTaskFactory after |thread_posting_tasks| is destroyed.
worker_pool_->WaitForAllWorkersIdleForTesting();
}
TEST_P(TaskSchedulerWorkerPoolImplTestParam, PostTasksWithOneAvailableWorker) {
// Post blocking tasks to keep all workers busy except one until |event| is
// signaled. Use different factories so that tasks are added to different
// sequences and can run simultaneously when the execution mode is SEQUENCED.
WaitableEvent event;
std::vector<std::unique_ptr<test::TestTaskFactory>> blocked_task_factories;
for (size_t i = 0; i < (kMaxTasks - 1); ++i) {
blocked_task_factories.push_back(std::make_unique<test::TestTaskFactory>(
CreateTaskRunnerWithExecutionMode(worker_pool_.get(), GetParam()),
GetParam()));
EXPECT_TRUE(blocked_task_factories.back()->PostTask(
PostNestedTask::NO,
BindOnce(&WaitWithoutBlockingObserver, Unretained(&event))));
blocked_task_factories.back()->WaitForAllTasksToRun();
}
// Post |kNumTasksPostedPerThread| tasks that should all run despite the fact
// that only one worker in |worker_pool_| isn't busy.
test::TestTaskFactory short_task_factory(
CreateTaskRunnerWithExecutionMode(worker_pool_.get(), GetParam()),
GetParam());
for (size_t i = 0; i < kNumTasksPostedPerThread; ++i)
EXPECT_TRUE(short_task_factory.PostTask(PostNestedTask::NO, Closure()));
short_task_factory.WaitForAllTasksToRun();
// Release tasks waiting on |event|.
event.Signal();
// Wait until all workers are idle to be sure that no task accesses
// its TestTaskFactory after it is destroyed.
worker_pool_->WaitForAllWorkersIdleForTesting();
}
TEST_P(TaskSchedulerWorkerPoolImplTestParam, Saturate) {
// Verify that it is possible to have |kMaxTasks| tasks/sequences running
// simultaneously. Use different factories so that the blocking tasks are
// added to different sequences and can run simultaneously when the execution
// mode is SEQUENCED.
WaitableEvent event;
std::vector<std::unique_ptr<test::TestTaskFactory>> factories;
for (size_t i = 0; i < kMaxTasks; ++i) {
factories.push_back(std::make_unique<test::TestTaskFactory>(
CreateTaskRunnerWithExecutionMode(worker_pool_.get(), GetParam()),
GetParam()));
EXPECT_TRUE(factories.back()->PostTask(
PostNestedTask::NO,
BindOnce(&WaitWithoutBlockingObserver, Unretained(&event))));
factories.back()->WaitForAllTasksToRun();
}
// Release tasks waiting on |event|.
event.Signal();
// Wait until all workers are idle to be sure that no task accesses
// its TestTaskFactory after it is destroyed.
worker_pool_->WaitForAllWorkersIdleForTesting();
}
#if defined(OS_WIN)
TEST_P(TaskSchedulerWorkerPoolImplTestParam, NoEnvironment) {
// Verify that COM is not initialized in a SchedulerWorkerPoolImpl initialized
// with SchedulerWorkerPoolImpl::WorkerEnvironment::NONE.
scoped_refptr<TaskRunner> task_runner =
CreateTaskRunnerWithExecutionMode(worker_pool_.get(), GetParam());
WaitableEvent task_running;
task_runner->PostTask(
FROM_HERE, BindOnce(
[](WaitableEvent* task_running) {
win::AssertComApartmentType(win::ComApartmentType::NONE);
task_running->Signal();
},
&task_running));
task_running.Wait();
worker_pool_->WaitForAllWorkersIdleForTesting();
}
#endif // defined(OS_WIN)
INSTANTIATE_TEST_CASE_P(Parallel,
TaskSchedulerWorkerPoolImplTestParam,
::testing::Values(test::ExecutionMode::PARALLEL));
INSTANTIATE_TEST_CASE_P(Sequenced,
TaskSchedulerWorkerPoolImplTestParam,
::testing::Values(test::ExecutionMode::SEQUENCED));
#if defined(OS_WIN)
namespace {
class TaskSchedulerWorkerPoolImplTestCOMMTAParam
: public TaskSchedulerWorkerPoolImplTestBase,
public testing::TestWithParam<test::ExecutionMode> {
protected:
TaskSchedulerWorkerPoolImplTestCOMMTAParam() = default;
void SetUp() override { TaskSchedulerWorkerPoolImplTestBase::CommonSetUp(); }
void TearDown() override {
TaskSchedulerWorkerPoolImplTestBase::CommonTearDown();
}
private:
void StartWorkerPool(TimeDelta suggested_reclaim_time,
size_t max_tasks) override {
ASSERT_TRUE(worker_pool_);
worker_pool_->Start(
SchedulerWorkerPoolParams(max_tasks, suggested_reclaim_time), max_tasks,
service_thread_.task_runner(), nullptr,
SchedulerWorkerPoolImpl::WorkerEnvironment::COM_MTA);
}
DISALLOW_COPY_AND_ASSIGN(TaskSchedulerWorkerPoolImplTestCOMMTAParam);
};
} // namespace
TEST_P(TaskSchedulerWorkerPoolImplTestCOMMTAParam, COMMTAInitialized) {
// Verify that SchedulerWorkerPoolImpl workers have a COM MTA available.
scoped_refptr<TaskRunner> task_runner =
CreateTaskRunnerWithExecutionMode(worker_pool_.get(), GetParam());
WaitableEvent task_running;
task_runner->PostTask(
FROM_HERE, BindOnce(
[](WaitableEvent* task_running) {
win::AssertComApartmentType(win::ComApartmentType::MTA);
task_running->Signal();
},
&task_running));
task_running.Wait();
worker_pool_->WaitForAllWorkersIdleForTesting();
}
INSTANTIATE_TEST_CASE_P(Parallel,
TaskSchedulerWorkerPoolImplTestCOMMTAParam,
::testing::Values(test::ExecutionMode::PARALLEL));
INSTANTIATE_TEST_CASE_P(Sequenced,
TaskSchedulerWorkerPoolImplTestCOMMTAParam,
::testing::Values(test::ExecutionMode::SEQUENCED));
#endif // defined(OS_WIN)
namespace {
class TaskSchedulerWorkerPoolImplStartInBodyTest
: public TaskSchedulerWorkerPoolImplTest {
public:
void SetUp() override {
CreateWorkerPool();
// Let the test start the worker pool.
}
};
void TaskPostedBeforeStart(PlatformThreadRef* platform_thread_ref,
WaitableEvent* task_running,
WaitableEvent* barrier) {
*platform_thread_ref = PlatformThread::CurrentRef();
task_running->Signal();
WaitWithoutBlockingObserver(barrier);
}
} // namespace
// Verify that 2 tasks posted before Start() to a SchedulerWorkerPoolImpl with
// more than 2 workers run on different workers when Start() is called.
TEST_F(TaskSchedulerWorkerPoolImplStartInBodyTest, PostTasksBeforeStart) {
PlatformThreadRef task_1_thread_ref;
PlatformThreadRef task_2_thread_ref;
WaitableEvent task_1_running;
WaitableEvent task_2_running;
// This event is used to prevent a task from completing before the other task
// starts running. If that happened, both tasks could run on the same worker
// and this test couldn't verify that the correct number of workers were woken
// up.
WaitableEvent barrier;
worker_pool_->CreateTaskRunnerWithTraits({WithBaseSyncPrimitives()})
->PostTask(
FROM_HERE,
BindOnce(&TaskPostedBeforeStart, Unretained(&task_1_thread_ref),
Unretained(&task_1_running), Unretained(&barrier)));
worker_pool_->CreateTaskRunnerWithTraits({WithBaseSyncPrimitives()})
->PostTask(
FROM_HERE,
BindOnce(&TaskPostedBeforeStart, Unretained(&task_2_thread_ref),
Unretained(&task_2_running), Unretained(&barrier)));
// Workers should not be created and tasks should not run before the pool is
// started.
EXPECT_EQ(0U, worker_pool_->NumberOfWorkersForTesting());
EXPECT_FALSE(task_1_running.IsSignaled());
EXPECT_FALSE(task_2_running.IsSignaled());
StartWorkerPool(TimeDelta::Max(), kMaxTasks);
// Tasks should run shortly after the pool is started.
task_1_running.Wait();
task_2_running.Wait();
// Tasks should run on different threads.
EXPECT_NE(task_1_thread_ref, task_2_thread_ref);
barrier.Signal();
task_tracker_.FlushForTesting();
}
// Verify that posting many tasks before Start will cause the number of workers
// to grow to |max_tasks_| during Start.
TEST_F(TaskSchedulerWorkerPoolImplStartInBodyTest, PostManyTasks) {
scoped_refptr<TaskRunner> task_runner =
worker_pool_->CreateTaskRunnerWithTraits({WithBaseSyncPrimitives()});
constexpr size_t kNumTasksPosted = 2 * kMaxTasks;
for (size_t i = 0; i < kNumTasksPosted; ++i)
task_runner->PostTask(FROM_HERE, DoNothing());
EXPECT_EQ(0U, worker_pool_->NumberOfWorkersForTesting());
StartWorkerPool(TimeDelta::Max(), kMaxTasks);
ASSERT_GT(kNumTasksPosted, worker_pool_->GetMaxTasksForTesting());
EXPECT_EQ(kMaxTasks, worker_pool_->GetMaxTasksForTesting());
EXPECT_EQ(worker_pool_->NumberOfWorkersForTesting(),
worker_pool_->GetMaxTasksForTesting());
}
namespace {
constexpr size_t kMagicTlsValue = 42;
class TaskSchedulerWorkerPoolCheckTlsReuse
: public TaskSchedulerWorkerPoolImplTest {
public:
void SetTlsValueAndWait() {
slot_.Set(reinterpret_cast<void*>(kMagicTlsValue));
WaitWithoutBlockingObserver(&waiter_);
}
void CountZeroTlsValuesAndWait(WaitableEvent* count_waiter) {
if (!slot_.Get())
subtle::NoBarrier_AtomicIncrement(&zero_tls_values_, 1);
count_waiter->Signal();
WaitWithoutBlockingObserver(&waiter_);
}
protected:
TaskSchedulerWorkerPoolCheckTlsReuse() = default;
void SetUp() override {
CreateAndStartWorkerPool(kReclaimTimeForCleanupTests, kMaxTasks);
}
subtle::Atomic32 zero_tls_values_ = 0;
WaitableEvent waiter_;
private:
ThreadLocalStorage::Slot slot_;
DISALLOW_COPY_AND_ASSIGN(TaskSchedulerWorkerPoolCheckTlsReuse);
};
} // namespace
// Checks that at least one worker has been cleaned up by checking the TLS.
TEST_F(TaskSchedulerWorkerPoolCheckTlsReuse, CheckCleanupWorkers) {
// Saturate the workers and mark each worker's thread with a magic TLS value.
std::vector<std::unique_ptr<test::TestTaskFactory>> factories;
for (size_t i = 0; i < kMaxTasks; ++i) {
factories.push_back(std::make_unique<test::TestTaskFactory>(
worker_pool_->CreateTaskRunnerWithTraits({WithBaseSyncPrimitives()}),
test::ExecutionMode::PARALLEL));
ASSERT_TRUE(factories.back()->PostTask(
PostNestedTask::NO,
Bind(&TaskSchedulerWorkerPoolCheckTlsReuse::SetTlsValueAndWait,
Unretained(this))));
factories.back()->WaitForAllTasksToRun();
}
// Release tasks waiting on |waiter_|.
waiter_.Signal();
worker_pool_->WaitForAllWorkersIdleForTesting();
// All workers should be done running by now, so reset for the next phase.
waiter_.Reset();
// Wait for the worker pool to clean up at least one worker.
worker_pool_->WaitForWorkersCleanedUpForTesting(1U);
// Saturate and count the worker threads that do not have the magic TLS value.
// If the value is not there, that means we're at a new worker.
std::vector<std::unique_ptr<WaitableEvent>> count_waiters;
for (auto& factory : factories) {
count_waiters.push_back(std::make_unique<WaitableEvent>());
ASSERT_TRUE(factory->PostTask(
PostNestedTask::NO,
Bind(&TaskSchedulerWorkerPoolCheckTlsReuse::CountZeroTlsValuesAndWait,
Unretained(this), count_waiters.back().get())));
factory->WaitForAllTasksToRun();
}
// Wait for all counters to complete.
for (auto& count_waiter : count_waiters)
count_waiter->Wait();
EXPECT_GT(subtle::NoBarrier_Load(&zero_tls_values_), 0);
// Release tasks waiting on |waiter_|.
waiter_.Signal();
}
#if !defined(STARBOARD)
namespace {
class TaskSchedulerWorkerPoolHistogramTest
: public TaskSchedulerWorkerPoolImplTest {
public:
TaskSchedulerWorkerPoolHistogramTest() = default;
protected:
// Override SetUp() to allow every test case to initialize a worker pool with
// its own arguments.
void SetUp() override {}
// Floods |worker_pool_| with a single task each that blocks until
// |continue_event| is signaled. Every worker in the pool is blocked on
// |continue_event| when this method returns. Note: this helper can easily be
// generalized to be useful in other tests, but it's here for now because it's
// only used in a TaskSchedulerWorkerPoolHistogramTest at the moment.
void FloodPool(WaitableEvent* continue_event) {
ASSERT_FALSE(continue_event->IsSignaled());
auto task_runner =
worker_pool_->CreateTaskRunnerWithTraits({WithBaseSyncPrimitives()});
const auto max_tasks = worker_pool_->GetMaxTasksForTesting();
WaitableEvent workers_flooded;
RepeatingClosure all_workers_running_barrier = BarrierClosure(
max_tasks,
BindOnce(&WaitableEvent::Signal, Unretained(&workers_flooded)));
for (size_t i = 0; i < max_tasks; ++i) {
task_runner->PostTask(
FROM_HERE,
BindOnce(
[](OnceClosure on_running, WaitableEvent* continue_event) {
std::move(on_running).Run();
WaitWithoutBlockingObserver(continue_event);
},
all_workers_running_barrier, continue_event));
}
workers_flooded.Wait();
}
private:
std::unique_ptr<StatisticsRecorder> statistics_recorder_ =
StatisticsRecorder::CreateTemporaryForTesting();
DISALLOW_COPY_AND_ASSIGN(TaskSchedulerWorkerPoolHistogramTest);
};
} // namespace
TEST_F(TaskSchedulerWorkerPoolHistogramTest, NumTasksBetweenWaits) {
WaitableEvent event;
CreateAndStartWorkerPool(TimeDelta::Max(), kMaxTasks);
auto task_runner = worker_pool_->CreateSequencedTaskRunnerWithTraits(
{WithBaseSyncPrimitives()});
// Post a task.
task_runner->PostTask(
FROM_HERE, BindOnce(&WaitWithoutBlockingObserver, Unretained(&event)));
// Post 2 more tasks while the first task hasn't completed its execution. It
// is guaranteed that these tasks will run immediately after the first task,
// without allowing the worker to sleep.
task_runner->PostTask(FROM_HERE, DoNothing());
task_runner->PostTask(FROM_HERE, DoNothing());
// Allow tasks to run and wait until the SchedulerWorker is idle.
event.Signal();
worker_pool_->WaitForAllWorkersIdleForTesting();
// Wake up the SchedulerWorker that just became idle by posting a task and
// wait until it becomes idle again. The SchedulerWorker should record the
// TaskScheduler.NumTasksBetweenWaits.* histogram on wake up.
task_runner->PostTask(FROM_HERE, DoNothing());
worker_pool_->WaitForAllWorkersIdleForTesting();
// Verify that counts were recorded to the histogram as expected.
const auto* histogram = worker_pool_->num_tasks_between_waits_histogram();
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(0));
EXPECT_EQ(1, histogram->SnapshotSamples()->GetCount(3));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(10));
}
// Verifies that NumTasksBetweenWaits histogram is logged as expected across
// idle and cleanup periods.
TEST_F(TaskSchedulerWorkerPoolHistogramTest,
NumTasksBetweenWaitsWithIdlePeriodAndCleanup) {
WaitableEvent tasks_can_exit_event;
CreateAndStartWorkerPool(kReclaimTimeForCleanupTests, kMaxTasks);
WaitableEvent workers_continue;
FloodPool(&workers_continue);
const auto* histogram = worker_pool_->num_tasks_between_waits_histogram();
// NumTasksBetweenWaits shouldn't be logged until idle.
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(0));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(1));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(10));
// Make all workers go idle.
workers_continue.Signal();
worker_pool_->WaitForAllWorkersIdleForTesting();
// All workers should have reported a single hit in the "1" bucket per the the
// histogram being reported when going idle and each worker having processed
// precisely 1 task per the controlled flooding logic above.
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(0));
EXPECT_EQ(static_cast<int>(kMaxTasks),
histogram->SnapshotSamples()->GetCount(1));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(10));
worker_pool_->WaitForWorkersCleanedUpForTesting(kMaxTasks - 1);
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(0));
EXPECT_EQ(static_cast<int>(kMaxTasks),
histogram->SnapshotSamples()->GetCount(1));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(10));
// Flooding the pool once again (without letting any workers go idle)
// shouldn't affect the counts either.
workers_continue.Reset();
FloodPool(&workers_continue);
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(0));
EXPECT_EQ(static_cast<int>(kMaxTasks),
histogram->SnapshotSamples()->GetCount(1));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(10));
workers_continue.Signal();
worker_pool_->WaitForAllWorkersIdleForTesting();
}
TEST_F(TaskSchedulerWorkerPoolHistogramTest, NumTasksBeforeCleanup) {
CreateWorkerPool();
auto histogrammed_thread_task_runner =
worker_pool_->CreateSequencedTaskRunnerWithTraits(
{WithBaseSyncPrimitives()});
// Post 3 tasks and hold the thread for idle thread stack ordering.
// This test assumes |histogrammed_thread_task_runner| gets assigned the same
// thread for each of its tasks.
PlatformThreadRef thread_ref;
histogrammed_thread_task_runner->PostTask(
FROM_HERE, BindOnce(
[](PlatformThreadRef* thread_ref) {
ASSERT_TRUE(thread_ref);
*thread_ref = PlatformThread::CurrentRef();
},
Unretained(&thread_ref)));
histogrammed_thread_task_runner->PostTask(
FROM_HERE, BindOnce(
[](PlatformThreadRef* thread_ref) {
ASSERT_FALSE(thread_ref->is_null());
EXPECT_EQ(*thread_ref, PlatformThread::CurrentRef());
},
Unretained(&thread_ref)));
WaitableEvent cleanup_thread_running;
WaitableEvent cleanup_thread_continue;
histogrammed_thread_task_runner->PostTask(
FROM_HERE,
BindOnce(
[](PlatformThreadRef* thread_ref,
WaitableEvent* cleanup_thread_running,
WaitableEvent* cleanup_thread_continue) {
ASSERT_FALSE(thread_ref->is_null());
EXPECT_EQ(*thread_ref, PlatformThread::CurrentRef());
cleanup_thread_running->Signal();
WaitWithoutBlockingObserver(cleanup_thread_continue);
},
Unretained(&thread_ref), Unretained(&cleanup_thread_running),
Unretained(&cleanup_thread_continue)));
// Start the worker pool with 2 workers, to avoid depending on the scheduler's
// logic to always keep one extra idle worker.
//
// The pool is started after the 3 initial tasks have been posted to ensure
// that they are scheduled on the same worker. If the tasks could run as they
// are posted, there would be a chance that:
// 1. Worker #1: Runs a tasks and empties the sequence, without adding
// itself to the idle stack yet.
// 2. Posting thread: Posts another task to the now empty sequence. Wakes
// up a new worker, since worker #1 isn't on the idle
// stack yet.
// 3: Worker #2: Runs the tasks, violating the expectation that the 3
// initial tasks run on the same worker.
constexpr size_t kTwoWorkers = 2;
StartWorkerPool(kReclaimTimeForCleanupTests, kTwoWorkers);
// Wait until the 3rd task is scheduled.
cleanup_thread_running.Wait();
// To allow the SchedulerWorker associated with
// |histogrammed_thread_task_runner| to cleanup, make sure it isn't on top of
// the idle stack by waking up another SchedulerWorker via
// |task_runner_for_top_idle|. |histogrammed_thread_task_runner| should
// release and go idle first and then |task_runner_for_top_idle| should
// release and go idle. This allows the SchedulerWorker associated with
// |histogrammed_thread_task_runner| to cleanup.
WaitableEvent top_idle_thread_running;
WaitableEvent top_idle_thread_continue;
auto task_runner_for_top_idle =
worker_pool_->CreateSequencedTaskRunnerWithTraits(
{WithBaseSyncPrimitives()});
task_runner_for_top_idle->PostTask(
FROM_HERE, BindOnce(
[](PlatformThreadRef thread_ref,
WaitableEvent* top_idle_thread_running,
WaitableEvent* top_idle_thread_continue) {
ASSERT_FALSE(thread_ref.is_null());
EXPECT_NE(thread_ref, PlatformThread::CurrentRef())
<< "Worker reused. Worker will not cleanup and the "
"histogram value will be wrong.";
top_idle_thread_running->Signal();
WaitWithoutBlockingObserver(top_idle_thread_continue);
},
thread_ref, Unretained(&top_idle_thread_running),
Unretained(&top_idle_thread_continue)));
top_idle_thread_running.Wait();
EXPECT_EQ(0U, worker_pool_->NumberOfIdleWorkersForTesting());
cleanup_thread_continue.Signal();
// Wait for the cleanup thread to also become idle.
worker_pool_->WaitForWorkersIdleForTesting(1U);
top_idle_thread_continue.Signal();
// Allow the thread processing the |histogrammed_thread_task_runner| work to
// cleanup.
worker_pool_->WaitForWorkersCleanedUpForTesting(1U);
// Verify that counts were recorded to the histogram as expected.
const auto* histogram = worker_pool_->num_tasks_before_detach_histogram();
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(0));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(1));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(2));
EXPECT_EQ(1, histogram->SnapshotSamples()->GetCount(3));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(4));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(5));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(6));
EXPECT_EQ(0, histogram->SnapshotSamples()->GetCount(10));
}
#endif // !defined(STARBOARD)
namespace {
class TaskSchedulerWorkerPoolStandbyPolicyTest
: public TaskSchedulerWorkerPoolImplTestBase,
public testing::Test {
public:
TaskSchedulerWorkerPoolStandbyPolicyTest() = default;
void SetUp() override {
TaskSchedulerWorkerPoolImplTestBase::CommonSetUp(
kReclaimTimeForCleanupTests);
}
void TearDown() override {
TaskSchedulerWorkerPoolImplTestBase::CommonTearDown();
}
private:
DISALLOW_COPY_AND_ASSIGN(TaskSchedulerWorkerPoolStandbyPolicyTest);
};
} // namespace
TEST_F(TaskSchedulerWorkerPoolStandbyPolicyTest, InitOne) {
EXPECT_EQ(1U, worker_pool_->NumberOfWorkersForTesting());
}
// Verify that the SchedulerWorkerPoolImpl keeps at least one idle standby
// thread, capacity permitting.
TEST_F(TaskSchedulerWorkerPoolStandbyPolicyTest, VerifyStandbyThread) {
auto task_runner =
worker_pool_->CreateTaskRunnerWithTraits({WithBaseSyncPrimitives()});
WaitableEvent thread_running(WaitableEvent::ResetPolicy::AUTOMATIC);
WaitableEvent threads_continue;
RepeatingClosure thread_blocker = BindLambdaForTesting([&]() {
thread_running.Signal();
WaitWithoutBlockingObserver(&threads_continue);
});
// There should be one idle thread until we reach capacity
for (size_t i = 0; i < kMaxTasks; ++i) {
EXPECT_EQ(i + 1, worker_pool_->NumberOfWorkersForTesting());
task_runner->PostTask(FROM_HERE, thread_blocker);
thread_running.Wait();
}
// There should not be an extra idle thread if it means going above capacity
EXPECT_EQ(kMaxTasks, worker_pool_->NumberOfWorkersForTesting());
threads_continue.Signal();
// Wait long enough for all but one worker to clean up.
worker_pool_->WaitForWorkersCleanedUpForTesting(kMaxTasks - 1);
EXPECT_EQ(1U, worker_pool_->NumberOfWorkersForTesting());
// Give extra time for a worker to cleanup : none should as the pool is
// expected to keep a worker ready regardless of how long it was idle for.
PlatformThread::Sleep(kReclaimTimeForCleanupTests);
EXPECT_EQ(1U, worker_pool_->NumberOfWorkersForTesting());
}
// Verify that being "the" idle thread counts as being active (i.e. won't be
// reclaimed even if not on top of the idle stack when reclaim timeout expires).
// Regression test for https://crbug.com/847501.
TEST_F(TaskSchedulerWorkerPoolStandbyPolicyTest,
InAndOutStandbyThreadIsActive) {
auto sequenced_task_runner =
worker_pool_->CreateSequencedTaskRunnerWithTraits({});
WaitableEvent timer_started;
RepeatingTimer recurring_task;
sequenced_task_runner->PostTask(
FROM_HERE, BindLambdaForTesting([&]() {
recurring_task.Start(FROM_HERE, kReclaimTimeForCleanupTests / 2,
DoNothing());
timer_started.Signal();
}));
timer_started.Wait();
// Running a task should have brought up a new standby thread.
EXPECT_EQ(2U, worker_pool_->NumberOfWorkersForTesting());
// Give extra time for a worker to cleanup : none should as the two workers
// are both considered "active" per the timer ticking faster than the reclaim
// timeout.
PlatformThread::Sleep(kReclaimTimeForCleanupTests * 2);
EXPECT_EQ(2U, worker_pool_->NumberOfWorkersForTesting());
sequenced_task_runner->PostTask(
FROM_HERE, BindLambdaForTesting([&]() { recurring_task.Stop(); }));
// Stopping the recurring task should let the second worker be reclaimed per
// not being "the" standby thread for a full reclaim timeout.
worker_pool_->WaitForWorkersCleanedUpForTesting(1);
EXPECT_EQ(1U, worker_pool_->NumberOfWorkersForTesting());
}
// Verify that being "the" idle thread counts as being active but isn't sticky.
// Regression test for https://crbug.com/847501.
TEST_F(TaskSchedulerWorkerPoolStandbyPolicyTest, OnlyKeepActiveStandbyThreads) {
auto sequenced_task_runner =
worker_pool_->CreateSequencedTaskRunnerWithTraits({});
// Start this test like
// TaskSchedulerWorkerPoolStandbyPolicyTest.InAndOutStandbyThreadIsActive and
// give it some time to stabilize.
RepeatingTimer recurring_task;
sequenced_task_runner->PostTask(
FROM_HERE, BindLambdaForTesting([&]() {
recurring_task.Start(FROM_HERE, kReclaimTimeForCleanupTests / 2,
DoNothing());
}));
PlatformThread::Sleep(kReclaimTimeForCleanupTests * 2);
EXPECT_EQ(2U, worker_pool_->NumberOfWorkersForTesting());
// Then also flood the pool (cycling the top of the idle stack).
{
auto task_runner =
worker_pool_->CreateTaskRunnerWithTraits({WithBaseSyncPrimitives()});
WaitableEvent thread_running(WaitableEvent::ResetPolicy::AUTOMATIC);
WaitableEvent threads_continue;
RepeatingClosure thread_blocker = BindLambdaForTesting([&]() {
thread_running.Signal();
WaitWithoutBlockingObserver(&threads_continue);
});
for (size_t i = 0; i < kMaxTasks; ++i) {
task_runner->PostTask(FROM_HERE, thread_blocker);
thread_running.Wait();
}
EXPECT_EQ(kMaxTasks, worker_pool_->NumberOfWorkersForTesting());
threads_continue.Signal();
// Flush to ensure all references to |threads_continue| are gone before it
// goes out of scope.
task_tracker_.FlushForTesting();
}
// All workers should clean up but two (since the timer is still running).
worker_pool_->WaitForWorkersCleanedUpForTesting(kMaxTasks - 2);
EXPECT_EQ(2U, worker_pool_->NumberOfWorkersForTesting());
// Extra time shouldn't change this.
PlatformThread::Sleep(kReclaimTimeForCleanupTests * 2);
EXPECT_EQ(2U, worker_pool_->NumberOfWorkersForTesting());
// Stopping the timer should let the number of active threads go down to one.
sequenced_task_runner->PostTask(
FROM_HERE, BindLambdaForTesting([&]() { recurring_task.Stop(); }));
worker_pool_->WaitForWorkersCleanedUpForTesting(1);
EXPECT_EQ(1U, worker_pool_->NumberOfWorkersForTesting());
}
namespace {
enum class OptionalBlockingType {
NO_BLOCK,
MAY_BLOCK,
WILL_BLOCK,
};
struct NestedBlockingType {
NestedBlockingType(BlockingType first_in,
OptionalBlockingType second_in,
BlockingType behaves_as_in)
: first(first_in), second(second_in), behaves_as(behaves_as_in) {}
BlockingType first;
OptionalBlockingType second;
BlockingType behaves_as;
};
class NestedScopedBlockingCall {
public:
NestedScopedBlockingCall(const NestedBlockingType& nested_blocking_type)
: first_scoped_blocking_call_(nested_blocking_type.first),
second_scoped_blocking_call_(
nested_blocking_type.second == OptionalBlockingType::WILL_BLOCK
? std::make_unique<ScopedBlockingCall>(BlockingType::WILL_BLOCK)
: (nested_blocking_type.second ==
OptionalBlockingType::MAY_BLOCK
? std::make_unique<ScopedBlockingCall>(
BlockingType::MAY_BLOCK)
: nullptr)) {}
private:
ScopedBlockingCall first_scoped_blocking_call_;
std::unique_ptr<ScopedBlockingCall> second_scoped_blocking_call_;
DISALLOW_COPY_AND_ASSIGN(NestedScopedBlockingCall);
};
} // namespace
class TaskSchedulerWorkerPoolBlockingTest
: public TaskSchedulerWorkerPoolImplTestBase,
public testing::TestWithParam<NestedBlockingType> {
public:
TaskSchedulerWorkerPoolBlockingTest() = default;
static std::string ParamInfoToString(
::testing::TestParamInfo<NestedBlockingType> param_info) {
std::string str = param_info.param.first == BlockingType::MAY_BLOCK
? "MAY_BLOCK"
: "WILL_BLOCK";
if (param_info.param.second == OptionalBlockingType::MAY_BLOCK)
str += "_MAY_BLOCK";
else if (param_info.param.second == OptionalBlockingType::WILL_BLOCK)
str += "_WILL_BLOCK";
return str;
}
void SetUp() override {
TaskSchedulerWorkerPoolImplTestBase::CommonSetUp();
task_runner_ = worker_pool_->CreateTaskRunnerWithTraits(
{MayBlock(), WithBaseSyncPrimitives()});
}
void TearDown() override {
TaskSchedulerWorkerPoolImplTestBase::CommonTearDown();
}
protected:
// Saturates the worker pool with a task that first blocks, waits to be
// unblocked, then exits.
void SaturateWithBlockingTasks(
const NestedBlockingType& nested_blocking_type) {
ASSERT_FALSE(blocking_threads_running_.IsSignaled());
RepeatingClosure blocking_threads_running_closure = BarrierClosure(
kMaxTasks, BindOnce(&WaitableEvent::Signal,
Unretained(&blocking_threads_running_)));
for (size_t i = 0; i < kMaxTasks; ++i) {
task_runner_->PostTask(
FROM_HERE,
BindOnce(
[](Closure* blocking_threads_running_closure,
WaitableEvent* blocking_threads_continue_,
const NestedBlockingType& nested_blocking_type) {
NestedScopedBlockingCall nested_scoped_blocking_call(
nested_blocking_type);
blocking_threads_running_closure->Run();
WaitWithoutBlockingObserver(blocking_threads_continue_);
},
Unretained(&blocking_threads_running_closure),
Unretained(&blocking_threads_continue_), nested_blocking_type));
}
blocking_threads_running_.Wait();
}
// Returns how long we can expect a change to |max_tasks_| to occur
// after a task has become blocked.
TimeDelta GetMaxTasksChangeSleepTime() {
return std::max(SchedulerWorkerPoolImpl::kBlockedWorkersPollPeriod,
worker_pool_->MayBlockThreshold()) +
TestTimeouts::tiny_timeout();
}
// Waits indefinitely, until |worker_pool_|'s max tasks increases to
// |expected_max_tasks|.
void ExpectMaxTasksIncreasesTo(size_t expected_max_tasks) {
size_t max_tasks = worker_pool_->GetMaxTasksForTesting();
while (max_tasks != expected_max_tasks) {
PlatformThread::Sleep(GetMaxTasksChangeSleepTime());
size_t new_max_tasks = worker_pool_->GetMaxTasksForTesting();
ASSERT_GE(new_max_tasks, max_tasks);
max_tasks = new_max_tasks;
}
}
// Unblocks tasks posted by SaturateWithBlockingTasks().
void UnblockTasks() { blocking_threads_continue_.Signal(); }
scoped_refptr<TaskRunner> task_runner_;
private:
WaitableEvent blocking_threads_running_;
WaitableEvent blocking_threads_continue_;
DISALLOW_COPY_AND_ASSIGN(TaskSchedulerWorkerPoolBlockingTest);
};
// Verify that BlockingScopeEntered() causes max tasks to increase and creates a
// worker if needed. Also verify that BlockingScopeExited() decreases max tasks
// after an increase.
TEST_P(TaskSchedulerWorkerPoolBlockingTest, ThreadBlockedUnblocked) {
ASSERT_EQ(worker_pool_->GetMaxTasksForTesting(), kMaxTasks);
SaturateWithBlockingTasks(GetParam());
if (GetParam().behaves_as == BlockingType::MAY_BLOCK)
ExpectMaxTasksIncreasesTo(2 * kMaxTasks);
// A range of possible number of workers is accepted because of
// crbug.com/757897.
EXPECT_GE(worker_pool_->NumberOfWorkersForTesting(), kMaxTasks + 1);
EXPECT_LE(worker_pool_->NumberOfWorkersForTesting(), 2 * kMaxTasks);
EXPECT_EQ(worker_pool_->GetMaxTasksForTesting(), 2 * kMaxTasks);
UnblockTasks();
task_tracker_.FlushForTesting();
EXPECT_EQ(worker_pool_->GetMaxTasksForTesting(), kMaxTasks);
}
// Verify that tasks posted in a saturated pool before a ScopedBlockingCall will
// execute after ScopedBlockingCall is instantiated.
TEST_P(TaskSchedulerWorkerPoolBlockingTest, PostBeforeBlocking) {
WaitableEvent thread_running(WaitableEvent::ResetPolicy::AUTOMATIC);
WaitableEvent thread_can_block;
WaitableEvent threads_continue;
for (size_t i = 0; i < kMaxTasks; ++i) {
task_runner_->PostTask(
FROM_HERE,
BindOnce(
[](const NestedBlockingType& nested_blocking_type,
WaitableEvent* thread_running, WaitableEvent* thread_can_block,
WaitableEvent* threads_continue) {
thread_running->Signal();
WaitWithoutBlockingObserver(thread_can_block);
NestedScopedBlockingCall nested_scoped_blocking_call(
nested_blocking_type);
WaitWithoutBlockingObserver(threads_continue);
},
GetParam(), Unretained(&thread_running),
Unretained(&thread_can_block), Unretained(&threads_continue)));
thread_running.Wait();
}
// All workers should be occupied and the pool should be saturated. Workers
// have not entered ScopedBlockingCall yet.
EXPECT_EQ(worker_pool_->NumberOfWorkersForTesting(), kMaxTasks);
EXPECT_EQ(worker_pool_->GetMaxTasksForTesting(), kMaxTasks);
WaitableEvent extra_threads_running;
WaitableEvent extra_threads_continue;
RepeatingClosure extra_threads_running_barrier = BarrierClosure(
kMaxTasks,
BindOnce(&WaitableEvent::Signal, Unretained(&extra_threads_running)));
for (size_t i = 0; i < kMaxTasks; ++i) {
task_runner_->PostTask(
FROM_HERE, BindOnce(
[](Closure* extra_threads_running_barrier,
WaitableEvent* extra_threads_continue) {
extra_threads_running_barrier->Run();
WaitWithoutBlockingObserver(extra_threads_continue);
},
Unretained(&extra_threads_running_barrier),
Unretained(&extra_threads_continue)));
}
// Allow tasks to enter ScopedBlockingCall. Workers should be created for the
// tasks we just posted.
thread_can_block.Signal();
if (GetParam().behaves_as == BlockingType::MAY_BLOCK)
ExpectMaxTasksIncreasesTo(2 * kMaxTasks);
// Should not block forever.
extra_threads_running.Wait();
EXPECT_EQ(worker_pool_->NumberOfWorkersForTesting(), 2 * kMaxTasks);
extra_threads_continue.Signal();
threads_continue.Signal();
task_tracker_.FlushForTesting();
}
// Verify that workers become idle when the pool is over-capacity and that
// those workers do no work.
TEST_P(TaskSchedulerWorkerPoolBlockingTest, WorkersIdleWhenOverCapacity) {
ASSERT_EQ(worker_pool_->GetMaxTasksForTesting(), kMaxTasks);
SaturateWithBlockingTasks(GetParam());
if (GetParam().behaves_as == BlockingType::MAY_BLOCK)
ExpectMaxTasksIncreasesTo(2 * kMaxTasks);
EXPECT_EQ(worker_pool_->GetMaxTasksForTesting(), 2 * kMaxTasks);
// A range of possible number of workers is accepted because of
// crbug.com/757897.
EXPECT_GE(worker_pool_->NumberOfWorkersForTesting(), kMaxTasks + 1);
EXPECT_LE(worker_pool_->NumberOfWorkersForTesting(), 2 * kMaxTasks);
WaitableEvent threads_running;
WaitableEvent threads_continue;
RepeatingClosure threads_running_barrier = BarrierClosure(
kMaxTasks,
BindOnce(&WaitableEvent::Signal, Unretained(&threads_running)));
// Posting these tasks should cause new workers to be created.
for (size_t i = 0; i < kMaxTasks; ++i) {
auto callback = BindOnce(
[](Closure* threads_running_barrier, WaitableEvent* threads_continue) {
threads_running_barrier->Run();
WaitWithoutBlockingObserver(threads_continue);
},
Unretained(&threads_running_barrier), Unretained(&threads_continue));
task_runner_->PostTask(FROM_HERE, std::move(callback));
}
threads_running.Wait();
ASSERT_EQ(worker_pool_->NumberOfIdleWorkersForTesting(), 0U);
EXPECT_EQ(worker_pool_->NumberOfWorkersForTesting(), 2 * kMaxTasks);
AtomicFlag is_exiting;
// These tasks should not get executed until after other tasks become
// unblocked.
for (size_t i = 0; i < kMaxTasks; ++i) {
task_runner_->PostTask(FROM_HERE, BindOnce(
[](AtomicFlag* is_exiting) {
EXPECT_TRUE(is_exiting->IsSet());
},
Unretained(&is_exiting)));
}
// The original |kMaxTasks| will finish their tasks after being
// unblocked. There will be work in the work queue, but the pool should now
// be over-capacity and workers will become idle.
UnblockTasks();
worker_pool_->WaitForWorkersIdleForTesting(kMaxTasks);
EXPECT_EQ(worker_pool_->NumberOfIdleWorkersForTesting(), kMaxTasks);
// Posting more tasks should not cause workers idle from the pool being over
// capacity to begin doing work.
for (size_t i = 0; i < kMaxTasks; ++i) {
task_runner_->PostTask(FROM_HERE, BindOnce(
[](AtomicFlag* is_exiting) {
EXPECT_TRUE(is_exiting->IsSet());
},
Unretained(&is_exiting)));
}
// Give time for those idle workers to possibly do work (which should not
// happen).
PlatformThread::Sleep(TestTimeouts::tiny_timeout());
is_exiting.Set();
// Unblocks the new workers.
threads_continue.Signal();
task_tracker_.FlushForTesting();
}
INSTANTIATE_TEST_CASE_P(
,
TaskSchedulerWorkerPoolBlockingTest,
::testing::Values(NestedBlockingType(BlockingType::MAY_BLOCK,
OptionalBlockingType::NO_BLOCK,
BlockingType::MAY_BLOCK),
NestedBlockingType(BlockingType::WILL_BLOCK,
OptionalBlockingType::NO_BLOCK,
BlockingType::WILL_BLOCK),
NestedBlockingType(BlockingType::MAY_BLOCK,
OptionalBlockingType::WILL_BLOCK,
BlockingType::WILL_BLOCK),
NestedBlockingType(BlockingType::WILL_BLOCK,
OptionalBlockingType::MAY_BLOCK,
BlockingType::WILL_BLOCK)),
TaskSchedulerWorkerPoolBlockingTest::ParamInfoToString);
// Verify that if a thread enters the scope of a MAY_BLOCK ScopedBlockingCall,
// but exits the scope before the MayBlockThreshold() is reached, that the max
// tasks does not increase.
TEST_F(TaskSchedulerWorkerPoolBlockingTest, ThreadBlockUnblockPremature) {
ASSERT_EQ(worker_pool_->GetMaxTasksForTesting(), kMaxTasks);
TimeDelta max_tasks_change_sleep = GetMaxTasksChangeSleepTime();
worker_pool_->MaximizeMayBlockThresholdForTesting();
SaturateWithBlockingTasks(NestedBlockingType(BlockingType::MAY_BLOCK,
OptionalBlockingType::NO_BLOCK,
BlockingType::MAY_BLOCK));
PlatformThread::Sleep(max_tasks_change_sleep);
EXPECT_EQ(worker_pool_->NumberOfWorkersForTesting(), kMaxTasks);
EXPECT_EQ(worker_pool_->GetMaxTasksForTesting(), kMaxTasks);
UnblockTasks();
task_tracker_.FlushForTesting();
EXPECT_EQ(worker_pool_->GetMaxTasksForTesting(), kMaxTasks);
}
// Verify that if max tasks is incremented because of a MAY_BLOCK
// ScopedBlockingCall, it isn't incremented again when there is a nested
// WILL_BLOCK ScopedBlockingCall.
TEST_F(TaskSchedulerWorkerPoolBlockingTest,
MayBlockIncreaseCapacityNestedWillBlock) {
ASSERT_EQ(worker_pool_->GetMaxTasksForTesting(), kMaxTasks);
auto task_runner = worker_pool_->CreateTaskRunnerWithTraits(
{MayBlock(), WithBaseSyncPrimitives()});
WaitableEvent can_return;
// Saturate the pool so that a MAY_BLOCK ScopedBlockingCall would increment
// the max tasks.
for (size_t i = 0; i < kMaxTasks - 1; ++i) {
task_runner->PostTask(FROM_HERE, BindOnce(&WaitWithoutBlockingObserver,
Unretained(&can_return)));
}
WaitableEvent can_instantiate_will_block;
WaitableEvent did_instantiate_will_block;
// Post a task that instantiates a MAY_BLOCK ScopedBlockingCall.
task_runner->PostTask(
FROM_HERE,
BindOnce(
[](WaitableEvent* can_instantiate_will_block,
WaitableEvent* did_instantiate_will_block,
WaitableEvent* can_return) {
ScopedBlockingCall may_block(BlockingType::MAY_BLOCK);
WaitWithoutBlockingObserver(can_instantiate_will_block);
ScopedBlockingCall will_block(BlockingType::WILL_BLOCK);
did_instantiate_will_block->Signal();
WaitWithoutBlockingObserver(can_return);
},
Unretained(&can_instantiate_will_block),
Unretained(&did_instantiate_will_block), Unretained(&can_return)));
// After a short delay, max tasks should be incremented.
ExpectMaxTasksIncreasesTo(kMaxTasks + 1);
// Wait until the task instantiates a WILL_BLOCK ScopedBlockingCall.
can_instantiate_will_block.Signal();
did_instantiate_will_block.Wait();
// Max tasks shouldn't be incremented again.
EXPECT_EQ(kMaxTasks + 1, worker_pool_->GetMaxTasksForTesting());
// Tear down.
can_return.Signal();
task_tracker_.FlushForTesting();
EXPECT_EQ(worker_pool_->GetMaxTasksForTesting(), kMaxTasks);
}
// Verify that workers that become idle due to the pool being over capacity will
// eventually cleanup.
TEST(TaskSchedulerWorkerPoolOverCapacityTest, VerifyCleanup) {
constexpr size_t kLocalMaxTasks = 3;
std::unique_ptr<StatisticsRecorder> recorder_for_testing =
StatisticsRecorder::CreateTemporaryForTesting();
TaskTracker task_tracker("Test");
DelayedTaskManager delayed_task_manager;
scoped_refptr<TaskRunner> service_thread_task_runner =
MakeRefCounted<TestSimpleTaskRunner>();
delayed_task_manager.Start(service_thread_task_runner);
SchedulerWorkerPoolImpl worker_pool(
"OverCapacityTestWorkerPool", "A", ThreadPriority::NORMAL,
task_tracker.GetTrackedRef(), &delayed_task_manager);
worker_pool.Start(
SchedulerWorkerPoolParams(kLocalMaxTasks, kReclaimTimeForCleanupTests),
kLocalMaxTasks, service_thread_task_runner, nullptr,
SchedulerWorkerPoolImpl::WorkerEnvironment::NONE);
scoped_refptr<TaskRunner> task_runner =
worker_pool.CreateTaskRunnerWithTraits(
{MayBlock(), WithBaseSyncPrimitives()});
WaitableEvent threads_running;
WaitableEvent threads_continue;
RepeatingClosure threads_running_barrier = BarrierClosure(
kLocalMaxTasks,
BindOnce(&WaitableEvent::Signal, Unretained(&threads_running)));
WaitableEvent blocked_call_continue;
RepeatingClosure closure = BindRepeating(
[](Closure* threads_running_barrier, WaitableEvent* threads_continue,
WaitableEvent* blocked_call_continue) {
threads_running_barrier->Run();
{
ScopedBlockingCall scoped_blocking_call(BlockingType::WILL_BLOCK);
WaitWithoutBlockingObserver(blocked_call_continue);
}
WaitWithoutBlockingObserver(threads_continue);
},
Unretained(&threads_running_barrier), Unretained(&threads_continue),
Unretained(&blocked_call_continue));
for (size_t i = 0; i < kLocalMaxTasks; ++i)
task_runner->PostTask(FROM_HERE, closure);
threads_running.Wait();
WaitableEvent extra_threads_running;
WaitableEvent extra_threads_continue;
RepeatingClosure extra_threads_running_barrier = BarrierClosure(
kLocalMaxTasks,
BindOnce(&WaitableEvent::Signal, Unretained(&extra_threads_running)));
// These tasks should run on the new threads from increasing max tasks.
for (size_t i = 0; i < kLocalMaxTasks; ++i) {
task_runner->PostTask(
FROM_HERE, BindOnce(
[](Closure* extra_threads_running_barrier,
WaitableEvent* extra_threads_continue) {
extra_threads_running_barrier->Run();
WaitWithoutBlockingObserver(extra_threads_continue);
},
Unretained(&extra_threads_running_barrier),
Unretained(&extra_threads_continue)));
}
extra_threads_running.Wait();
ASSERT_EQ(kLocalMaxTasks * 2, worker_pool.NumberOfWorkersForTesting());
EXPECT_EQ(kLocalMaxTasks * 2, worker_pool.GetMaxTasksForTesting());
blocked_call_continue.Signal();
extra_threads_continue.Signal();
// Periodically post tasks to ensure that posting tasks does not prevent
// workers that are idle due to the pool being over capacity from cleaning up.
for (int i = 0; i < 16; ++i) {
task_runner->PostDelayedTask(FROM_HERE, DoNothing(),
kReclaimTimeForCleanupTests * i * 0.5);
}
// Note: one worker above capacity will not get cleaned up since it's on the
// top of the idle stack.
worker_pool.WaitForWorkersCleanedUpForTesting(kLocalMaxTasks - 1);
EXPECT_EQ(kLocalMaxTasks + 1, worker_pool.NumberOfWorkersForTesting());
threads_continue.Signal();
worker_pool.JoinForTesting();
}
// Verify that the maximum number of workers is 256 and that hitting the max
// leaves the pool in a valid state with regards to max tasks.
TEST_F(TaskSchedulerWorkerPoolBlockingTest, MaximumWorkersTest) {
#ifdef STARBOARD
const size_t kMaxNumberOfWorkers = kSbMaxThreads;
#else
constexpr size_t kMaxNumberOfWorkers = 256;
#endif
constexpr size_t kNumExtraTasks = 10;
WaitableEvent early_blocking_threads_running;
RepeatingClosure early_threads_barrier_closure =
BarrierClosure(kMaxNumberOfWorkers,
BindOnce(&WaitableEvent::Signal,
Unretained(&early_blocking_threads_running)));
WaitableEvent early_threads_finished;
RepeatingClosure early_threads_finished_barrier = BarrierClosure(
kMaxNumberOfWorkers,
BindOnce(&WaitableEvent::Signal, Unretained(&early_threads_finished)));
WaitableEvent early_release_threads_continue;
// Post ScopedBlockingCall tasks to hit the worker cap.
for (size_t i = 0; i < kMaxNumberOfWorkers; ++i) {
task_runner_->PostTask(
FROM_HERE,
BindOnce(
[](Closure* early_threads_barrier_closure,
WaitableEvent* early_release_threads_continue,
Closure* early_threads_finished) {
{
ScopedBlockingCall scoped_blocking_call(
BlockingType::WILL_BLOCK);
early_threads_barrier_closure->Run();
WaitWithoutBlockingObserver(early_release_threads_continue);
}
early_threads_finished->Run();
},
Unretained(&early_threads_barrier_closure),
Unretained(&early_release_threads_continue),
Unretained(&early_threads_finished_barrier)));
}
early_blocking_threads_running.Wait();
EXPECT_EQ(worker_pool_->GetMaxTasksForTesting(),
kMaxTasks + kMaxNumberOfWorkers);
WaitableEvent late_release_thread_contine;
WaitableEvent late_blocking_threads_running;
RepeatingClosure late_threads_barrier_closure = BarrierClosure(
kNumExtraTasks, BindOnce(&WaitableEvent::Signal,
Unretained(&late_blocking_threads_running)));
// Posts additional tasks. Note: we should already have |kMaxNumberOfWorkers|
// tasks running. These tasks should not be able to get executed yet as
// the pool is already at its max worker cap.
for (size_t i = 0; i < kNumExtraTasks; ++i) {
task_runner_->PostTask(
FROM_HERE,
BindOnce(
[](Closure* late_threads_barrier_closure,
WaitableEvent* late_release_thread_contine) {
ScopedBlockingCall scoped_blocking_call(BlockingType::WILL_BLOCK);
late_threads_barrier_closure->Run();
WaitWithoutBlockingObserver(late_release_thread_contine);
},
Unretained(&late_threads_barrier_closure),
Unretained(&late_release_thread_contine)));
}
// Give time to see if we exceed the max number of workers.
PlatformThread::Sleep(TestTimeouts::tiny_timeout());
EXPECT_LE(worker_pool_->NumberOfWorkersForTesting(), kMaxNumberOfWorkers);
early_release_threads_continue.Signal();
early_threads_finished.Wait();
late_blocking_threads_running.Wait();
WaitableEvent final_tasks_running;
WaitableEvent final_tasks_continue;
RepeatingClosure final_tasks_running_barrier = BarrierClosure(
kMaxTasks,
BindOnce(&WaitableEvent::Signal, Unretained(&final_tasks_running)));
// Verify that we are still able to saturate the pool.
for (size_t i = 0; i < kMaxTasks; ++i) {
task_runner_->PostTask(
FROM_HERE,
BindOnce(
[](Closure* closure, WaitableEvent* final_tasks_continue) {
closure->Run();
WaitWithoutBlockingObserver(final_tasks_continue);
},
Unretained(&final_tasks_running_barrier),
Unretained(&final_tasks_continue)));
}
final_tasks_running.Wait();
EXPECT_EQ(worker_pool_->GetMaxTasksForTesting(), kMaxTasks + kNumExtraTasks);
late_release_thread_contine.Signal();
final_tasks_continue.Signal();
task_tracker_.FlushForTesting();
}
// Verify that the maximum number of best-effort tasks that can run concurrently
// is honored.
TEST_F(TaskSchedulerWorkerPoolImplStartInBodyTest, MaxBestEffortTasks) {
constexpr int kMaxBestEffortTasks = kMaxTasks / 2;
worker_pool_->Start(
SchedulerWorkerPoolParams(kMaxTasks, base::TimeDelta::Max()),
kMaxBestEffortTasks, service_thread_.task_runner(), nullptr,
SchedulerWorkerPoolImpl::WorkerEnvironment::NONE);
const scoped_refptr<TaskRunner> foreground_runner =
worker_pool_->CreateTaskRunnerWithTraits({MayBlock()});
const scoped_refptr<TaskRunner> background_runner =
worker_pool_->CreateTaskRunnerWithTraits(
{TaskPriority::BEST_EFFORT, MayBlock()});
// It should be possible to have |kMaxBestEffortTasks|
// TaskPriority::BEST_EFFORT tasks running concurrently.
WaitableEvent best_effort_tasks_running;
WaitableEvent unblock_best_effort_tasks;
RepeatingClosure best_effort_tasks_running_barrier = BarrierClosure(
kMaxBestEffortTasks,
BindOnce(&WaitableEvent::Signal, Unretained(&best_effort_tasks_running)));
for (int i = 0; i < kMaxBestEffortTasks; ++i) {
background_runner->PostTask(
FROM_HERE, base::BindLambdaForTesting([&]() {
best_effort_tasks_running_barrier.Run();
WaitWithoutBlockingObserver(&unblock_best_effort_tasks);
}));
}
best_effort_tasks_running.Wait();
// No more TaskPriority::BEST_EFFORT task should run.
AtomicFlag extra_best_effort_task_can_run;
WaitableEvent extra_best_effort_task_running;
background_runner->PostTask(
FROM_HERE, base::BindLambdaForTesting([&]() {
EXPECT_TRUE(extra_best_effort_task_can_run.IsSet());
extra_best_effort_task_running.Signal();
}));
// An extra foreground task should be able to run.
WaitableEvent foreground_task_running;
foreground_runner->PostTask(
FROM_HERE, base::BindOnce(&WaitableEvent::Signal,
Unretained(&foreground_task_running)));
foreground_task_running.Wait();
// Completion of the TaskPriority::BEST_EFFORT tasks should allow the extra
// TaskPriority::BEST_EFFORT task to run.
extra_best_effort_task_can_run.Set();
unblock_best_effort_tasks.Signal();
extra_best_effort_task_running.Wait();
// Tear down.
task_tracker_.FlushForTesting();
}
namespace {
class TaskSchedulerWorkerPoolBlockingCallAndMaxBestEffortTasksTest
: public TaskSchedulerWorkerPoolImplTestBase,
public testing::TestWithParam<BlockingType> {
public:
static constexpr int kMaxBestEffortTasks = kMaxTasks / 2;
TaskSchedulerWorkerPoolBlockingCallAndMaxBestEffortTasksTest() = default;
void SetUp() override {
CreateWorkerPool();
worker_pool_->Start(
SchedulerWorkerPoolParams(kMaxTasks, base::TimeDelta::Max()),
kMaxBestEffortTasks, service_thread_.task_runner(), nullptr,
SchedulerWorkerPoolImpl::WorkerEnvironment::NONE);
}
void TearDown() override {
TaskSchedulerWorkerPoolImplTestBase::CommonTearDown();
}
private:
DISALLOW_COPY_AND_ASSIGN(
TaskSchedulerWorkerPoolBlockingCallAndMaxBestEffortTasksTest);
};
} // namespace
TEST_P(TaskSchedulerWorkerPoolBlockingCallAndMaxBestEffortTasksTest,
BlockingCallAndMaxBestEffortTasksTest) {
const scoped_refptr<TaskRunner> background_runner =
worker_pool_->CreateTaskRunnerWithTraits(
{TaskPriority::BEST_EFFORT, MayBlock()});
// Post |kMaxBestEffortTasks| TaskPriority::BEST_EFFORT tasks that block in a
// ScopedBlockingCall.
WaitableEvent blocking_best_effort_tasks_running;
WaitableEvent unblock_blocking_best_effort_tasks;
RepeatingClosure blocking_best_effort_tasks_running_barrier =
BarrierClosure(kMaxBestEffortTasks,
BindOnce(&WaitableEvent::Signal,
Unretained(&blocking_best_effort_tasks_running)));
for (int i = 0; i < kMaxBestEffortTasks; ++i) {
background_runner->PostTask(
FROM_HERE, base::BindLambdaForTesting([&]() {
blocking_best_effort_tasks_running_barrier.Run();
ScopedBlockingCall scoped_blocking_call(GetParam());
WaitWithoutBlockingObserver(&unblock_blocking_best_effort_tasks);
}));
}
blocking_best_effort_tasks_running.Wait();
// Post an extra |kMaxBestEffortTasks| TaskPriority::BEST_EFFORT tasks. They
// should be able to run, because the existing TaskPriority::BEST_EFFORT tasks
// are blocked within a ScopedBlockingCall.
//
// Note: We block the tasks until they have all started running to make sure
// that it is possible to run an extra |kMaxBestEffortTasks| concurrently.
WaitableEvent best_effort_tasks_running;
WaitableEvent unblock_best_effort_tasks;
RepeatingClosure best_effort_tasks_running_barrier = BarrierClosure(
kMaxBestEffortTasks,
BindOnce(&WaitableEvent::Signal, Unretained(&best_effort_tasks_running)));
for (int i = 0; i < kMaxBestEffortTasks; ++i) {
background_runner->PostTask(
FROM_HERE, base::BindLambdaForTesting([&]() {
best_effort_tasks_running_barrier.Run();
WaitWithoutBlockingObserver(&unblock_best_effort_tasks);
}));
}
best_effort_tasks_running.Wait();
// Unblock all tasks and tear down.
unblock_blocking_best_effort_tasks.Signal();
unblock_best_effort_tasks.Signal();
task_tracker_.FlushForTesting();
}
INSTANTIATE_TEST_CASE_P(
MayBlock,
TaskSchedulerWorkerPoolBlockingCallAndMaxBestEffortTasksTest,
::testing::Values(BlockingType::MAY_BLOCK));
INSTANTIATE_TEST_CASE_P(
WillBlock,
TaskSchedulerWorkerPoolBlockingCallAndMaxBestEffortTasksTest,
::testing::Values(BlockingType::WILL_BLOCK));
// Verify that worker detachement doesn't race with worker cleanup, regression
// test for https://crbug.com/810464.
TEST_F(TaskSchedulerWorkerPoolImplStartInBodyTest, RacyCleanup) {
#ifdef STARBOARD
const size_t kLocalMaxTasks = kSbMaxThreads;
#else
#if defined(OS_FUCHSIA)
// Fuchsia + QEMU doesn't deal well with *many* threads being
// created/destroyed at once: https://crbug.com/816575.
constexpr size_t kLocalMaxTasks = 16;
#else // defined(OS_FUCHSIA)
constexpr size_t kLocalMaxTasks = 256;
#endif // defined(OS_FUCHSIA)
#endif // STARBOARD
constexpr TimeDelta kReclaimTimeForRacyCleanupTest =
TimeDelta::FromMilliseconds(10);
worker_pool_->Start(
SchedulerWorkerPoolParams(kLocalMaxTasks, kReclaimTimeForRacyCleanupTest),
kLocalMaxTasks, service_thread_.task_runner(), nullptr,
SchedulerWorkerPoolImpl::WorkerEnvironment::NONE);
scoped_refptr<TaskRunner> task_runner =
worker_pool_->CreateTaskRunnerWithTraits({WithBaseSyncPrimitives()});
WaitableEvent threads_running;
WaitableEvent unblock_threads;
RepeatingClosure threads_running_barrier = BarrierClosure(
kLocalMaxTasks,
BindOnce(&WaitableEvent::Signal, Unretained(&threads_running)));
for (size_t i = 0; i < kLocalMaxTasks; ++i) {
task_runner->PostTask(
FROM_HERE,
BindOnce(
[](OnceClosure on_running, WaitableEvent* unblock_threads) {
std::move(on_running).Run();
WaitWithoutBlockingObserver(unblock_threads);
},
threads_running_barrier, Unretained(&unblock_threads)));
}
// Wait for all workers to be ready and release them all at once.
threads_running.Wait();
unblock_threads.Signal();
// Sleep to wakeup precisely when all workers are going to try to cleanup per
// being idle.
PlatformThread::Sleep(kReclaimTimeForRacyCleanupTest);
worker_pool_->JoinForTesting();
// Unwinding this test will be racy if worker cleanup can race with
// SchedulerWorkerPoolImpl destruction : https://crbug.com/810464.
worker_pool_.reset();
}
TEST_P(TaskSchedulerWorkerPoolImplTestParam, RecordNumWorkersHistogram) {
HistogramTester tester;
worker_pool_->RecordNumWorkersHistogram();
EXPECT_FALSE(
tester.GetAllSamples("TaskScheduler.NumWorkers.TestWorkerPoolPool")
.empty());
}
} // namespace internal
} // namespace base