blob: 6bd6730cde34752a3e7db50ee8d2f153379013da [file] [log] [blame]
// Copyright 2013 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/message_loop/message_loop.h"
#include <algorithm>
#include "base/bind.h"
#include "base/compiler_specific.h"
#include "base/lazy_instance.h"
#include "base/logging.h"
#include "base/memory/scoped_ptr.h"
#include "base/message_loop/message_pump_default.h"
#include "base/metrics/histogram.h"
#include "base/metrics/statistics_recorder.h"
#include "base/run_loop.h"
#include "base/third_party/dynamic_annotations/dynamic_annotations.h"
#include "base/thread_task_runner_handle.h"
#include "base/threading/thread_local.h"
#include "base/time/time.h"
#include "base/trace_event/trace_event.h"
#include "base/tracked_objects.h"
#if defined(OS_MACOSX)
#include "base/message_loop/message_pump_mac.h"
#endif
#if defined(OS_POSIX) && !defined(OS_IOS)
#include "base/message_loop/message_pump_libevent.h"
#endif
#if defined(OS_ANDROID)
#include "base/message_loop/message_pump_android.h"
#endif
#if defined(USE_GLIB)
#include "base/message_loop/message_pump_glib.h"
#endif
namespace base {
namespace {
// A lazily created thread local storage for quick access to a thread's message
// loop, if one exists. This should be safe and free of static constructors.
LazyInstance<base::ThreadLocalPointer<MessageLoop> >::Leaky lazy_tls_ptr =
LAZY_INSTANCE_INITIALIZER;
// Logical events for Histogram profiling. Run with -message-loop-histogrammer
// to get an accounting of messages and actions taken on each thread.
const int kTaskRunEvent = 0x1;
#if !defined(OS_NACL)
const int kTimerEvent = 0x2;
// Provide range of message IDs for use in histogramming and debug display.
const int kLeastNonZeroMessageId = 1;
const int kMaxMessageId = 1099;
const int kNumberOfDistinctMessagesDisplayed = 1100;
// Provide a macro that takes an expression (such as a constant, or macro
// constant) and creates a pair to initalize an array of pairs. In this case,
// our pair consists of the expressions value, and the "stringized" version
// of the expression (i.e., the exrpression put in quotes). For example, if
// we have:
// #define FOO 2
// #define BAR 5
// then the following:
// VALUE_TO_NUMBER_AND_NAME(FOO + BAR)
// will expand to:
// {7, "FOO + BAR"}
// We use the resulting array as an argument to our histogram, which reads the
// number as a bucket identifier, and proceeds to use the corresponding name
// in the pair (i.e., the quoted string) when printing out a histogram.
#define VALUE_TO_NUMBER_AND_NAME(name) {name, #name},
const LinearHistogram::DescriptionPair event_descriptions_[] = {
// Provide some pretty print capability in our histogram for our internal
// messages.
// A few events we handle (kindred to messages), and used to profile actions.
VALUE_TO_NUMBER_AND_NAME(kTaskRunEvent)
VALUE_TO_NUMBER_AND_NAME(kTimerEvent)
{-1, NULL} // The list must be null terminated, per API to histogram.
};
#endif // !defined(OS_NACL)
bool enable_histogrammer_ = false;
MessageLoop::MessagePumpFactory* message_pump_for_ui_factory_ = NULL;
#if defined(OS_IOS)
typedef MessagePumpIOSForIO MessagePumpForIO;
#elif defined(OS_NACL_SFI)
typedef MessagePumpDefault MessagePumpForIO;
#elif defined(OS_POSIX)
typedef MessagePumpLibevent MessagePumpForIO;
#endif
#if !defined(OS_NACL_SFI)
MessagePumpForIO* ToPumpIO(MessagePump* pump) {
return static_cast<MessagePumpForIO*>(pump);
}
#endif // !defined(OS_NACL_SFI)
scoped_ptr<MessagePump> ReturnPump(scoped_ptr<MessagePump> pump) {
return pump;
}
} // namespace
//------------------------------------------------------------------------------
MessageLoop::TaskObserver::TaskObserver() {
}
MessageLoop::TaskObserver::~TaskObserver() {
}
MessageLoop::DestructionObserver::~DestructionObserver() {
}
//------------------------------------------------------------------------------
MessageLoop::MessageLoop(Type type)
: MessageLoop(type, MessagePumpFactoryCallback()) {
BindToCurrentThread();
}
MessageLoop::MessageLoop(scoped_ptr<MessagePump> pump)
: MessageLoop(TYPE_CUSTOM, Bind(&ReturnPump, Passed(&pump))) {
BindToCurrentThread();
}
MessageLoop::~MessageLoop() {
// current() could be NULL if this message loop is destructed before it is
// bound to a thread.
DCHECK(current() == this || !current());
// iOS just attaches to the loop, it doesn't Run it.
// TODO(stuartmorgan): Consider wiring up a Detach().
#if !defined(OS_IOS)
DCHECK(!run_loop_);
#endif
#if defined(OS_WIN)
if (in_high_res_mode_)
Time::ActivateHighResolutionTimer(false);
#endif
// Clean up any unprocessed tasks, but take care: deleting a task could
// result in the addition of more tasks (e.g., via DeleteSoon). We set a
// limit on the number of times we will allow a deleted task to generate more
// tasks. Normally, we should only pass through this loop once or twice. If
// we end up hitting the loop limit, then it is probably due to one task that
// is being stubborn. Inspect the queues to see who is left.
bool did_work;
for (int i = 0; i < 100; ++i) {
DeletePendingTasks();
ReloadWorkQueue();
// If we end up with empty queues, then break out of the loop.
did_work = DeletePendingTasks();
if (!did_work)
break;
}
DCHECK(!did_work);
// Let interested parties have one last shot at accessing this.
FOR_EACH_OBSERVER(DestructionObserver, destruction_observers_,
WillDestroyCurrentMessageLoop());
thread_task_runner_handle_.reset();
// Tell the incoming queue that we are dying.
incoming_task_queue_->WillDestroyCurrentMessageLoop();
incoming_task_queue_ = NULL;
unbound_task_runner_ = NULL;
task_runner_ = NULL;
// OK, now make it so that no one can find us.
lazy_tls_ptr.Pointer()->Set(NULL);
}
// static
MessageLoop* MessageLoop::current() {
// TODO(darin): sadly, we cannot enable this yet since people call us even
// when they have no intention of using us.
// DCHECK(loop) << "Ouch, did you forget to initialize me?";
return lazy_tls_ptr.Pointer()->Get();
}
// static
void MessageLoop::EnableHistogrammer(bool enable) {
enable_histogrammer_ = enable;
}
// static
bool MessageLoop::InitMessagePumpForUIFactory(MessagePumpFactory* factory) {
if (message_pump_for_ui_factory_)
return false;
message_pump_for_ui_factory_ = factory;
return true;
}
// static
scoped_ptr<MessagePump> MessageLoop::CreateMessagePumpForType(Type type) {
// TODO(rvargas): Get rid of the OS guards.
#if defined(USE_GLIB) && !defined(OS_NACL)
typedef MessagePumpGlib MessagePumpForUI;
#elif defined(OS_LINUX) && !defined(OS_NACL)
typedef MessagePumpLibevent MessagePumpForUI;
#endif
#if defined(OS_IOS) || defined(OS_MACOSX)
#define MESSAGE_PUMP_UI scoped_ptr<MessagePump>(MessagePumpMac::Create())
#elif defined(OS_NACL)
// Currently NaCl doesn't have a UI MessageLoop.
// TODO(abarth): Figure out if we need this.
#define MESSAGE_PUMP_UI scoped_ptr<MessagePump>()
#else
#define MESSAGE_PUMP_UI scoped_ptr<MessagePump>(new MessagePumpForUI())
#endif
#if defined(OS_MACOSX)
// Use an OS native runloop on Mac to support timer coalescing.
#define MESSAGE_PUMP_DEFAULT \
scoped_ptr<MessagePump>(new MessagePumpCFRunLoop())
#else
#define MESSAGE_PUMP_DEFAULT scoped_ptr<MessagePump>(new MessagePumpDefault())
#endif
if (type == MessageLoop::TYPE_UI) {
if (message_pump_for_ui_factory_)
return message_pump_for_ui_factory_();
return MESSAGE_PUMP_UI;
}
if (type == MessageLoop::TYPE_IO)
return scoped_ptr<MessagePump>(new MessagePumpForIO());
#if defined(OS_ANDROID)
if (type == MessageLoop::TYPE_JAVA)
return scoped_ptr<MessagePump>(new MessagePumpForUI());
#endif
DCHECK_EQ(MessageLoop::TYPE_DEFAULT, type);
return MESSAGE_PUMP_DEFAULT;
}
void MessageLoop::AddDestructionObserver(
DestructionObserver* destruction_observer) {
DCHECK_EQ(this, current());
destruction_observers_.AddObserver(destruction_observer);
}
void MessageLoop::RemoveDestructionObserver(
DestructionObserver* destruction_observer) {
DCHECK_EQ(this, current());
destruction_observers_.RemoveObserver(destruction_observer);
}
void MessageLoop::PostTask(
const tracked_objects::Location& from_here,
const Closure& task) {
task_runner_->PostTask(from_here, task);
}
void MessageLoop::PostDelayedTask(
const tracked_objects::Location& from_here,
const Closure& task,
TimeDelta delay) {
task_runner_->PostDelayedTask(from_here, task, delay);
}
void MessageLoop::PostNonNestableTask(
const tracked_objects::Location& from_here,
const Closure& task) {
task_runner_->PostNonNestableTask(from_here, task);
}
void MessageLoop::PostNonNestableDelayedTask(
const tracked_objects::Location& from_here,
const Closure& task,
TimeDelta delay) {
task_runner_->PostNonNestableDelayedTask(from_here, task, delay);
}
void MessageLoop::Run() {
DCHECK(pump_);
RunLoop run_loop;
run_loop.Run();
}
void MessageLoop::RunUntilIdle() {
DCHECK(pump_);
RunLoop run_loop;
run_loop.RunUntilIdle();
}
void MessageLoop::QuitWhenIdle() {
DCHECK_EQ(this, current());
if (run_loop_) {
run_loop_->quit_when_idle_received_ = true;
} else {
NOTREACHED() << "Must be inside Run to call Quit";
}
}
void MessageLoop::QuitNow() {
DCHECK_EQ(this, current());
if (run_loop_) {
pump_->Quit();
} else {
NOTREACHED() << "Must be inside Run to call Quit";
}
}
bool MessageLoop::IsType(Type type) const {
return type_ == type;
}
static void QuitCurrentWhenIdle() {
MessageLoop::current()->QuitWhenIdle();
}
// static
Closure MessageLoop::QuitWhenIdleClosure() {
return Bind(&QuitCurrentWhenIdle);
}
void MessageLoop::SetNestableTasksAllowed(bool allowed) {
if (allowed) {
// Kick the native pump just in case we enter a OS-driven nested message
// loop.
pump_->ScheduleWork();
}
nestable_tasks_allowed_ = allowed;
}
bool MessageLoop::NestableTasksAllowed() const {
return nestable_tasks_allowed_;
}
bool MessageLoop::IsNested() {
return run_loop_->run_depth_ > 1;
}
void MessageLoop::AddTaskObserver(TaskObserver* task_observer) {
DCHECK_EQ(this, current());
task_observers_.AddObserver(task_observer);
}
void MessageLoop::RemoveTaskObserver(TaskObserver* task_observer) {
DCHECK_EQ(this, current());
task_observers_.RemoveObserver(task_observer);
}
bool MessageLoop::is_running() const {
DCHECK_EQ(this, current());
return run_loop_ != NULL;
}
bool MessageLoop::HasHighResolutionTasks() {
return incoming_task_queue_->HasHighResolutionTasks();
}
bool MessageLoop::IsIdleForTesting() {
// We only check the imcoming queue|, since we don't want to lock the work
// queue.
return incoming_task_queue_->IsIdleForTesting();
}
//------------------------------------------------------------------------------
// static
scoped_ptr<MessageLoop> MessageLoop::CreateUnbound(
Type type, MessagePumpFactoryCallback pump_factory) {
return make_scoped_ptr(new MessageLoop(type, pump_factory));
}
MessageLoop::MessageLoop(Type type, MessagePumpFactoryCallback pump_factory)
: type_(type),
#if defined(OS_WIN)
pending_high_res_tasks_(0),
in_high_res_mode_(false),
#endif
nestable_tasks_allowed_(true),
#if defined(OS_WIN)
os_modal_loop_(false),
#endif // OS_WIN
pump_factory_(pump_factory),
message_histogram_(NULL),
run_loop_(NULL),
incoming_task_queue_(new internal::IncomingTaskQueue(this)),
unbound_task_runner_(
new internal::MessageLoopTaskRunner(incoming_task_queue_)),
task_runner_(unbound_task_runner_) {
// If type is TYPE_CUSTOM non-null pump_factory must be given.
DCHECK_EQ(type_ == TYPE_CUSTOM, !pump_factory_.is_null());
}
void MessageLoop::BindToCurrentThread() {
DCHECK(!pump_);
if (!pump_factory_.is_null())
pump_ = pump_factory_.Run();
else
pump_ = CreateMessagePumpForType(type_);
DCHECK(!current()) << "should only have one message loop per thread";
lazy_tls_ptr.Pointer()->Set(this);
incoming_task_queue_->StartScheduling();
unbound_task_runner_->BindToCurrentThread();
unbound_task_runner_ = nullptr;
SetThreadTaskRunnerHandle();
}
void MessageLoop::SetTaskRunner(
scoped_refptr<SingleThreadTaskRunner> task_runner) {
DCHECK_EQ(this, current());
DCHECK(task_runner->BelongsToCurrentThread());
DCHECK(!unbound_task_runner_);
task_runner_ = task_runner.Pass();
SetThreadTaskRunnerHandle();
}
void MessageLoop::SetThreadTaskRunnerHandle() {
DCHECK_EQ(this, current());
// Clear the previous thread task runner first because only one can exist at
// a time.
thread_task_runner_handle_.reset();
thread_task_runner_handle_.reset(new ThreadTaskRunnerHandle(task_runner_));
}
void MessageLoop::RunHandler() {
DCHECK_EQ(this, current());
StartHistogrammer();
#if defined(OS_WIN)
if (run_loop_->dispatcher_ && type() == TYPE_UI) {
static_cast<MessagePumpForUI*>(pump_.get())->
RunWithDispatcher(this, run_loop_->dispatcher_);
return;
}
#endif
pump_->Run(this);
}
bool MessageLoop::ProcessNextDelayedNonNestableTask() {
if (run_loop_->run_depth_ != 1)
return false;
if (deferred_non_nestable_work_queue_.empty())
return false;
PendingTask pending_task = deferred_non_nestable_work_queue_.front();
deferred_non_nestable_work_queue_.pop();
RunTask(pending_task);
return true;
}
void MessageLoop::RunTask(const PendingTask& pending_task) {
DCHECK(nestable_tasks_allowed_);
#if defined(OS_WIN)
if (pending_task.is_high_res) {
pending_high_res_tasks_--;
CHECK_GE(pending_high_res_tasks_, 0);
}
#endif
// Execute the task and assume the worst: It is probably not reentrant.
nestable_tasks_allowed_ = false;
HistogramEvent(kTaskRunEvent);
TRACE_TASK_EXECUTION("toplevel", pending_task);
FOR_EACH_OBSERVER(TaskObserver, task_observers_,
WillProcessTask(pending_task));
task_annotator_.RunTask("MessageLoop::PostTask", pending_task);
FOR_EACH_OBSERVER(TaskObserver, task_observers_,
DidProcessTask(pending_task));
nestable_tasks_allowed_ = true;
}
bool MessageLoop::DeferOrRunPendingTask(const PendingTask& pending_task) {
if (pending_task.nestable || run_loop_->run_depth_ == 1) {
RunTask(pending_task);
// Show that we ran a task (Note: a new one might arrive as a
// consequence!).
return true;
}
// We couldn't run the task now because we're in a nested message loop
// and the task isn't nestable.
deferred_non_nestable_work_queue_.push(pending_task);
return false;
}
void MessageLoop::AddToDelayedWorkQueue(const PendingTask& pending_task) {
// Move to the delayed work queue.
delayed_work_queue_.push(pending_task);
}
bool MessageLoop::DeletePendingTasks() {
bool did_work = !work_queue_.empty();
while (!work_queue_.empty()) {
PendingTask pending_task = work_queue_.front();
work_queue_.pop();
if (!pending_task.delayed_run_time.is_null()) {
// We want to delete delayed tasks in the same order in which they would
// normally be deleted in case of any funny dependencies between delayed
// tasks.
AddToDelayedWorkQueue(pending_task);
}
}
did_work |= !deferred_non_nestable_work_queue_.empty();
while (!deferred_non_nestable_work_queue_.empty()) {
deferred_non_nestable_work_queue_.pop();
}
did_work |= !delayed_work_queue_.empty();
// Historically, we always delete the task regardless of valgrind status. It's
// not completely clear why we want to leak them in the loops above. This
// code is replicating legacy behavior, and should not be considered
// absolutely "correct" behavior. See TODO above about deleting all tasks
// when it's safe.
while (!delayed_work_queue_.empty()) {
delayed_work_queue_.pop();
}
return did_work;
}
void MessageLoop::ReloadWorkQueue() {
// We can improve performance of our loading tasks from the incoming queue to
// |*work_queue| by waiting until the last minute (|*work_queue| is empty) to
// load. That reduces the number of locks-per-task significantly when our
// queues get large.
if (work_queue_.empty()) {
#if defined(OS_WIN)
pending_high_res_tasks_ +=
incoming_task_queue_->ReloadWorkQueue(&work_queue_);
#else
incoming_task_queue_->ReloadWorkQueue(&work_queue_);
#endif
}
}
void MessageLoop::ScheduleWork() {
pump_->ScheduleWork();
}
//------------------------------------------------------------------------------
// Method and data for histogramming events and actions taken by each instance
// on each thread.
void MessageLoop::StartHistogrammer() {
#if !defined(OS_NACL) // NaCl build has no metrics code.
if (enable_histogrammer_ && !message_histogram_
&& StatisticsRecorder::IsActive()) {
DCHECK(!thread_name_.empty());
message_histogram_ = LinearHistogram::FactoryGetWithRangeDescription(
"MsgLoop:" + thread_name_,
kLeastNonZeroMessageId, kMaxMessageId,
kNumberOfDistinctMessagesDisplayed,
message_histogram_->kHexRangePrintingFlag,
event_descriptions_);
}
#endif
}
void MessageLoop::HistogramEvent(int event) {
#if !defined(OS_NACL)
if (message_histogram_)
message_histogram_->Add(event);
#endif
}
bool MessageLoop::DoWork() {
if (!nestable_tasks_allowed_) {
// Task can't be executed right now.
return false;
}
for (;;) {
ReloadWorkQueue();
if (work_queue_.empty())
break;
// Execute oldest task.
do {
PendingTask pending_task = work_queue_.front();
work_queue_.pop();
if (!pending_task.delayed_run_time.is_null()) {
AddToDelayedWorkQueue(pending_task);
// If we changed the topmost task, then it is time to reschedule.
if (delayed_work_queue_.top().task.Equals(pending_task.task))
pump_->ScheduleDelayedWork(pending_task.delayed_run_time);
} else {
if (DeferOrRunPendingTask(pending_task))
return true;
}
} while (!work_queue_.empty());
}
// Nothing happened.
return false;
}
bool MessageLoop::DoDelayedWork(TimeTicks* next_delayed_work_time) {
if (!nestable_tasks_allowed_ || delayed_work_queue_.empty()) {
recent_time_ = *next_delayed_work_time = TimeTicks();
return false;
}
// When we "fall behind," there will be a lot of tasks in the delayed work
// queue that are ready to run. To increase efficiency when we fall behind,
// we will only call Time::Now() intermittently, and then process all tasks
// that are ready to run before calling it again. As a result, the more we
// fall behind (and have a lot of ready-to-run delayed tasks), the more
// efficient we'll be at handling the tasks.
TimeTicks next_run_time = delayed_work_queue_.top().delayed_run_time;
if (next_run_time > recent_time_) {
recent_time_ = TimeTicks::Now(); // Get a better view of Now();
if (next_run_time > recent_time_) {
*next_delayed_work_time = next_run_time;
return false;
}
}
PendingTask pending_task = delayed_work_queue_.top();
delayed_work_queue_.pop();
if (!delayed_work_queue_.empty())
*next_delayed_work_time = delayed_work_queue_.top().delayed_run_time;
return DeferOrRunPendingTask(pending_task);
}
bool MessageLoop::DoIdleWork() {
if (ProcessNextDelayedNonNestableTask())
return true;
if (run_loop_->quit_when_idle_received_)
pump_->Quit();
// When we return we will do a kernel wait for more tasks.
#if defined(OS_WIN)
// On Windows we activate the high resolution timer so that the wait
// _if_ triggered by the timer happens with good resolution. If we don't
// do this the default resolution is 15ms which might not be acceptable
// for some tasks.
bool high_res = pending_high_res_tasks_ > 0;
if (high_res != in_high_res_mode_) {
in_high_res_mode_ = high_res;
Time::ActivateHighResolutionTimer(in_high_res_mode_);
}
#endif
return false;
}
void MessageLoop::DeleteSoonInternal(const tracked_objects::Location& from_here,
void(*deleter)(const void*),
const void* object) {
PostNonNestableTask(from_here, Bind(deleter, object));
}
void MessageLoop::ReleaseSoonInternal(
const tracked_objects::Location& from_here,
void(*releaser)(const void*),
const void* object) {
PostNonNestableTask(from_here, Bind(releaser, object));
}
#if !defined(OS_NACL)
//------------------------------------------------------------------------------
// MessageLoopForUI
#if defined(OS_ANDROID)
void MessageLoopForUI::Start() {
// No Histogram support for UI message loop as it is managed by Java side
static_cast<MessagePumpForUI*>(pump_.get())->Start(this);
}
#endif
#if defined(OS_IOS)
void MessageLoopForUI::Attach() {
static_cast<MessagePumpUIApplication*>(pump_.get())->Attach(this);
}
#endif
#if defined(USE_OZONE) || (defined(USE_X11) && !defined(USE_GLIB))
bool MessageLoopForUI::WatchFileDescriptor(
int fd,
bool persistent,
MessagePumpLibevent::Mode mode,
MessagePumpLibevent::FileDescriptorWatcher *controller,
MessagePumpLibevent::Watcher *delegate) {
return static_cast<MessagePumpLibevent*>(pump_.get())->WatchFileDescriptor(
fd,
persistent,
mode,
controller,
delegate);
}
#endif
#endif // !defined(OS_NACL)
//------------------------------------------------------------------------------
// MessageLoopForIO
#if !defined(OS_NACL_SFI)
void MessageLoopForIO::AddIOObserver(
MessageLoopForIO::IOObserver* io_observer) {
ToPumpIO(pump_.get())->AddIOObserver(io_observer);
}
void MessageLoopForIO::RemoveIOObserver(
MessageLoopForIO::IOObserver* io_observer) {
ToPumpIO(pump_.get())->RemoveIOObserver(io_observer);
}
#if defined(OS_WIN)
void MessageLoopForIO::RegisterIOHandler(HANDLE file, IOHandler* handler) {
ToPumpIO(pump_.get())->RegisterIOHandler(file, handler);
}
bool MessageLoopForIO::RegisterJobObject(HANDLE job, IOHandler* handler) {
return ToPumpIO(pump_.get())->RegisterJobObject(job, handler);
}
bool MessageLoopForIO::WaitForIOCompletion(DWORD timeout, IOHandler* filter) {
return ToPumpIO(pump_.get())->WaitForIOCompletion(timeout, filter);
}
#elif defined(OS_POSIX)
bool MessageLoopForIO::WatchFileDescriptor(int fd,
bool persistent,
Mode mode,
FileDescriptorWatcher* controller,
Watcher* delegate) {
return ToPumpIO(pump_.get())->WatchFileDescriptor(
fd,
persistent,
mode,
controller,
delegate);
}
#endif
#endif // !defined(OS_NACL_SFI)
} // namespace base