mirror of
https://github.com/LadybirdBrowser/ladybird.git
synced 2024-11-26 17:40:27 +00:00
9601b516b8
This was used in exactly one place, to avoid sending multiple CustomEvents to the enqueuer thread in Audio::ConnectionToServer. Instead of this, we now just send a CustomEvent and wake the enqueuer thread. If it wakes up and has multiple CustomEvents, they get delivered and ignored in no time anyway. Since they only get ignored if there's no work to be done, this seems harmless.
925 lines
30 KiB
C++
925 lines
30 KiB
C++
/*
|
|
* Copyright (c) 2018-2023, Andreas Kling <kling@serenityos.org>
|
|
* Copyright (c) 2022, kleines Filmröllchen <malu.bertsch@gmail.com>
|
|
* Copyright (c) 2022, the SerenityOS developers.
|
|
*
|
|
* SPDX-License-Identifier: BSD-2-Clause
|
|
*/
|
|
|
|
#include <AK/Assertions.h>
|
|
#include <AK/Badge.h>
|
|
#include <AK/Debug.h>
|
|
#include <AK/Format.h>
|
|
#include <AK/IDAllocator.h>
|
|
#include <AK/JsonObject.h>
|
|
#include <AK/JsonValue.h>
|
|
#include <AK/NeverDestroyed.h>
|
|
#include <AK/Singleton.h>
|
|
#include <AK/TemporaryChange.h>
|
|
#include <AK/Time.h>
|
|
#include <LibCore/Event.h>
|
|
#include <LibCore/EventLoop.h>
|
|
#include <LibCore/LocalServer.h>
|
|
#include <LibCore/Notifier.h>
|
|
#include <LibCore/Object.h>
|
|
#include <LibCore/Promise.h>
|
|
#include <LibCore/SessionManagement.h>
|
|
#include <LibCore/Socket.h>
|
|
#include <LibThreading/Mutex.h>
|
|
#include <LibThreading/MutexProtected.h>
|
|
#include <errno.h>
|
|
#include <fcntl.h>
|
|
#include <signal.h>
|
|
#include <stdio.h>
|
|
#include <string.h>
|
|
#include <sys/select.h>
|
|
#include <sys/socket.h>
|
|
#include <sys/time.h>
|
|
#include <sys/types.h>
|
|
#include <time.h>
|
|
#include <unistd.h>
|
|
|
|
#ifdef AK_OS_SERENITY
|
|
# include <LibCore/Account.h>
|
|
|
|
extern bool s_global_initializers_ran;
|
|
#endif
|
|
|
|
namespace Core {
|
|
|
|
class InspectorServerConnection;
|
|
|
|
[[maybe_unused]] static bool connect_to_inspector_server();
|
|
|
|
struct EventLoopTimer {
|
|
int timer_id { 0 };
|
|
Time interval;
|
|
Time fire_time;
|
|
bool should_reload { false };
|
|
TimerShouldFireWhenNotVisible fire_when_not_visible { TimerShouldFireWhenNotVisible::No };
|
|
WeakPtr<Object> owner;
|
|
|
|
void reload(Time const& now);
|
|
bool has_expired(Time const& now) const;
|
|
};
|
|
|
|
struct EventLoop::Private {
|
|
Threading::Mutex lock;
|
|
};
|
|
|
|
static Threading::MutexProtected<NeverDestroyed<IDAllocator>> s_id_allocator;
|
|
static Threading::MutexProtected<RefPtr<InspectorServerConnection>> s_inspector_server_connection;
|
|
|
|
// Each thread has its own event loop stack, its own timers, notifiers and a wake pipe.
|
|
static thread_local Vector<EventLoop&>* s_event_loop_stack;
|
|
static thread_local HashMap<int, NonnullOwnPtr<EventLoopTimer>>* s_timers;
|
|
static thread_local HashTable<Notifier*>* s_notifiers;
|
|
// The wake pipe is both responsible for notifying us when someone calls wake(), as well as POSIX signals.
|
|
// While wake() pushes zero into the pipe, signal numbers (by defintion nonzero, see signal_numbers.h) are pushed into the pipe verbatim.
|
|
thread_local int EventLoop::s_wake_pipe_fds[2];
|
|
thread_local bool EventLoop::s_wake_pipe_initialized { false };
|
|
thread_local bool s_warned_promise_count { false };
|
|
|
|
void EventLoop::initialize_wake_pipes()
|
|
{
|
|
if (!s_wake_pipe_initialized) {
|
|
#if defined(SOCK_NONBLOCK)
|
|
int rc = pipe2(s_wake_pipe_fds, O_CLOEXEC);
|
|
#else
|
|
int rc = pipe(s_wake_pipe_fds);
|
|
fcntl(s_wake_pipe_fds[0], F_SETFD, FD_CLOEXEC);
|
|
fcntl(s_wake_pipe_fds[1], F_SETFD, FD_CLOEXEC);
|
|
|
|
#endif
|
|
VERIFY(rc == 0);
|
|
s_wake_pipe_initialized = true;
|
|
}
|
|
}
|
|
|
|
bool EventLoop::has_been_instantiated()
|
|
{
|
|
return s_event_loop_stack != nullptr && !s_event_loop_stack->is_empty();
|
|
}
|
|
|
|
class SignalHandlers : public RefCounted<SignalHandlers> {
|
|
AK_MAKE_NONCOPYABLE(SignalHandlers);
|
|
AK_MAKE_NONMOVABLE(SignalHandlers);
|
|
|
|
public:
|
|
SignalHandlers(int signo, void (*handle_signal)(int));
|
|
~SignalHandlers();
|
|
|
|
void dispatch();
|
|
int add(Function<void(int)>&& handler);
|
|
bool remove(int handler_id);
|
|
|
|
bool is_empty() const
|
|
{
|
|
if (m_calling_handlers) {
|
|
for (auto& handler : m_handlers_pending) {
|
|
if (handler.value)
|
|
return false; // an add is pending
|
|
}
|
|
}
|
|
return m_handlers.is_empty();
|
|
}
|
|
|
|
bool have(int handler_id) const
|
|
{
|
|
if (m_calling_handlers) {
|
|
auto it = m_handlers_pending.find(handler_id);
|
|
if (it != m_handlers_pending.end()) {
|
|
if (!it->value)
|
|
return false; // a deletion is pending
|
|
}
|
|
}
|
|
return m_handlers.contains(handler_id);
|
|
}
|
|
|
|
int m_signo;
|
|
void (*m_original_handler)(int); // TODO: can't use sighandler_t?
|
|
HashMap<int, Function<void(int)>> m_handlers;
|
|
HashMap<int, Function<void(int)>> m_handlers_pending;
|
|
bool m_calling_handlers { false };
|
|
};
|
|
|
|
struct SignalHandlersInfo {
|
|
HashMap<int, NonnullRefPtr<SignalHandlers>> signal_handlers;
|
|
int next_signal_id { 0 };
|
|
};
|
|
|
|
static Singleton<SignalHandlersInfo> s_signals;
|
|
template<bool create_if_null = true>
|
|
inline SignalHandlersInfo* signals_info()
|
|
{
|
|
return s_signals.ptr();
|
|
}
|
|
|
|
pid_t EventLoop::s_pid;
|
|
|
|
class InspectorServerConnection : public Object {
|
|
C_OBJECT(InspectorServerConnection)
|
|
private:
|
|
explicit InspectorServerConnection(NonnullOwnPtr<LocalSocket> socket)
|
|
: m_socket(move(socket))
|
|
, m_client_id(s_id_allocator.with_locked([](auto& allocator) {
|
|
return allocator->allocate();
|
|
}))
|
|
{
|
|
#ifdef AK_OS_SERENITY
|
|
m_socket->on_ready_to_read = [this] {
|
|
u32 length;
|
|
auto maybe_bytes_read = m_socket->read_some({ (u8*)&length, sizeof(length) });
|
|
if (maybe_bytes_read.is_error()) {
|
|
dbgln("InspectorServerConnection: Failed to read message length from inspector server connection: {}", maybe_bytes_read.error());
|
|
shutdown();
|
|
return;
|
|
}
|
|
|
|
auto bytes_read = maybe_bytes_read.release_value();
|
|
if (bytes_read.is_empty()) {
|
|
dbgln_if(EVENTLOOP_DEBUG, "RPC client disconnected");
|
|
shutdown();
|
|
return;
|
|
}
|
|
|
|
VERIFY(bytes_read.size() == sizeof(length));
|
|
|
|
auto request_buffer = ByteBuffer::create_uninitialized(length).release_value();
|
|
maybe_bytes_read = m_socket->read_some(request_buffer.bytes());
|
|
if (maybe_bytes_read.is_error()) {
|
|
dbgln("InspectorServerConnection: Failed to read message content from inspector server connection: {}", maybe_bytes_read.error());
|
|
shutdown();
|
|
return;
|
|
}
|
|
|
|
bytes_read = maybe_bytes_read.release_value();
|
|
|
|
auto request_json = JsonValue::from_string(request_buffer);
|
|
if (request_json.is_error() || !request_json.value().is_object()) {
|
|
dbgln("RPC client sent invalid request");
|
|
shutdown();
|
|
return;
|
|
}
|
|
|
|
handle_request(request_json.value().as_object());
|
|
};
|
|
#else
|
|
warnln("RPC Client constructed outside serenity, this is very likely a bug!");
|
|
#endif
|
|
}
|
|
virtual ~InspectorServerConnection() override
|
|
{
|
|
if (auto inspected_object = m_inspected_object.strong_ref())
|
|
inspected_object->decrement_inspector_count({});
|
|
}
|
|
|
|
public:
|
|
void send_response(JsonObject const& response)
|
|
{
|
|
auto serialized = response.to_deprecated_string();
|
|
auto bytes_to_send = serialized.bytes();
|
|
u32 length = bytes_to_send.size();
|
|
// FIXME: Propagate errors
|
|
MUST(m_socket->write_value(length));
|
|
while (!bytes_to_send.is_empty()) {
|
|
size_t bytes_sent = MUST(m_socket->write_some(bytes_to_send));
|
|
bytes_to_send = bytes_to_send.slice(bytes_sent);
|
|
}
|
|
}
|
|
|
|
void handle_request(JsonObject const& request)
|
|
{
|
|
auto type = request.get_deprecated_string("type"sv);
|
|
|
|
if (!type.has_value()) {
|
|
dbgln("RPC client sent request without type field");
|
|
return;
|
|
}
|
|
|
|
if (type == "Identify") {
|
|
JsonObject response;
|
|
response.set("type", type.value());
|
|
response.set("pid", getpid());
|
|
#ifdef AK_OS_SERENITY
|
|
char buffer[1024];
|
|
if (get_process_name(buffer, sizeof(buffer)) >= 0) {
|
|
response.set("process_name", buffer);
|
|
} else {
|
|
response.set("process_name", JsonValue());
|
|
}
|
|
#endif
|
|
send_response(response);
|
|
return;
|
|
}
|
|
|
|
if (type == "GetAllObjects") {
|
|
JsonObject response;
|
|
response.set("type", type.value());
|
|
JsonArray objects;
|
|
for (auto& object : Object::all_objects()) {
|
|
JsonObject json_object;
|
|
object.save_to(json_object);
|
|
objects.must_append(move(json_object));
|
|
}
|
|
response.set("objects", move(objects));
|
|
send_response(response);
|
|
return;
|
|
}
|
|
|
|
if (type == "SetInspectedObject") {
|
|
auto address = request.get_addr("address"sv);
|
|
for (auto& object : Object::all_objects()) {
|
|
if ((FlatPtr)&object == address) {
|
|
if (auto inspected_object = m_inspected_object.strong_ref())
|
|
inspected_object->decrement_inspector_count({});
|
|
m_inspected_object = object;
|
|
object.increment_inspector_count({});
|
|
break;
|
|
}
|
|
}
|
|
return;
|
|
}
|
|
|
|
if (type == "SetProperty") {
|
|
auto address = request.get_addr("address"sv);
|
|
for (auto& object : Object::all_objects()) {
|
|
if ((FlatPtr)&object == address) {
|
|
bool success = object.set_property(request.get_deprecated_string("name"sv).value(), request.get("value"sv).value());
|
|
JsonObject response;
|
|
response.set("type", "SetProperty");
|
|
response.set("success", success);
|
|
send_response(response);
|
|
break;
|
|
}
|
|
}
|
|
return;
|
|
}
|
|
|
|
if (type == "Disconnect") {
|
|
shutdown();
|
|
return;
|
|
}
|
|
}
|
|
|
|
void shutdown()
|
|
{
|
|
s_id_allocator.with_locked([this](auto& allocator) { allocator->deallocate(m_client_id); });
|
|
}
|
|
|
|
private:
|
|
NonnullOwnPtr<LocalSocket> m_socket;
|
|
WeakPtr<Object> m_inspected_object;
|
|
int m_client_id { -1 };
|
|
};
|
|
|
|
EventLoop::EventLoop([[maybe_unused]] MakeInspectable make_inspectable)
|
|
: m_wake_pipe_fds(&s_wake_pipe_fds)
|
|
, m_private(make<Private>())
|
|
{
|
|
#ifdef AK_OS_SERENITY
|
|
if (!s_global_initializers_ran) {
|
|
// NOTE: Trying to have an event loop as a global variable will lead to initialization-order fiascos,
|
|
// as the event loop constructor accesses and/or sets other global variables.
|
|
// Therefore, we crash the program before ASAN catches us.
|
|
// If you came here because of the assertion failure, please redesign your program to not have global event loops.
|
|
// The common practice is to initialize the main event loop in the main function, and if necessary,
|
|
// pass event loop references around or access them with EventLoop::with_main_locked() and EventLoop::current().
|
|
VERIFY_NOT_REACHED();
|
|
}
|
|
#endif
|
|
|
|
if (!s_event_loop_stack) {
|
|
s_event_loop_stack = new Vector<EventLoop&>;
|
|
s_timers = new HashMap<int, NonnullOwnPtr<EventLoopTimer>>;
|
|
s_notifiers = new HashTable<Notifier*>;
|
|
}
|
|
|
|
if (s_event_loop_stack->is_empty()) {
|
|
s_pid = getpid();
|
|
s_event_loop_stack->append(*this);
|
|
|
|
#ifdef AK_OS_SERENITY
|
|
if (getuid() != 0) {
|
|
if (getenv("MAKE_INSPECTABLE") == "1"sv)
|
|
make_inspectable = Core::EventLoop::MakeInspectable::Yes;
|
|
|
|
if (make_inspectable == MakeInspectable::Yes
|
|
&& !s_inspector_server_connection.with_locked([](auto inspector_server_connection) { return inspector_server_connection; })) {
|
|
if (!connect_to_inspector_server())
|
|
dbgln("Core::EventLoop: Failed to connect to InspectorServer");
|
|
}
|
|
}
|
|
#endif
|
|
}
|
|
|
|
initialize_wake_pipes();
|
|
|
|
dbgln_if(EVENTLOOP_DEBUG, "{} Core::EventLoop constructed :)", getpid());
|
|
}
|
|
|
|
EventLoop::~EventLoop()
|
|
{
|
|
if (!s_event_loop_stack->is_empty() && &s_event_loop_stack->last() == this)
|
|
s_event_loop_stack->take_last();
|
|
}
|
|
|
|
bool connect_to_inspector_server()
|
|
{
|
|
#ifdef AK_OS_SERENITY
|
|
auto maybe_path = SessionManagement::parse_path_with_sid("/tmp/session/%sid/portal/inspectables"sv);
|
|
if (maybe_path.is_error()) {
|
|
dbgln("connect_to_inspector_server: {}", maybe_path.error());
|
|
return false;
|
|
}
|
|
auto inspector_server_path = maybe_path.value();
|
|
auto maybe_socket = LocalSocket::connect(inspector_server_path, Socket::PreventSIGPIPE::Yes);
|
|
if (maybe_socket.is_error()) {
|
|
dbgln("connect_to_inspector_server: Failed to connect: {}", maybe_socket.error());
|
|
return false;
|
|
}
|
|
s_inspector_server_connection.with_locked([&](auto& inspector_server_connection) {
|
|
inspector_server_connection = InspectorServerConnection::construct(maybe_socket.release_value());
|
|
});
|
|
return true;
|
|
#else
|
|
VERIFY_NOT_REACHED();
|
|
#endif
|
|
}
|
|
|
|
#define VERIFY_EVENT_LOOP_INITIALIZED() \
|
|
do { \
|
|
if (!s_event_loop_stack) { \
|
|
warnln("EventLoop static API was called without prior EventLoop init!"); \
|
|
VERIFY_NOT_REACHED(); \
|
|
} \
|
|
} while (0)
|
|
|
|
EventLoop& EventLoop::current()
|
|
{
|
|
VERIFY_EVENT_LOOP_INITIALIZED();
|
|
return s_event_loop_stack->last();
|
|
}
|
|
|
|
void EventLoop::quit(int code)
|
|
{
|
|
dbgln_if(EVENTLOOP_DEBUG, "Core::EventLoop::quit({})", code);
|
|
m_exit_requested = true;
|
|
m_exit_code = code;
|
|
}
|
|
|
|
void EventLoop::unquit()
|
|
{
|
|
dbgln_if(EVENTLOOP_DEBUG, "Core::EventLoop::unquit()");
|
|
m_exit_requested = false;
|
|
m_exit_code = 0;
|
|
}
|
|
|
|
struct EventLoopPusher {
|
|
public:
|
|
EventLoopPusher(EventLoop& event_loop)
|
|
: m_event_loop(event_loop)
|
|
{
|
|
if (EventLoop::has_been_instantiated()) {
|
|
m_event_loop.take_pending_events_from(EventLoop::current());
|
|
s_event_loop_stack->append(event_loop);
|
|
}
|
|
}
|
|
~EventLoopPusher()
|
|
{
|
|
if (EventLoop::has_been_instantiated()) {
|
|
s_event_loop_stack->take_last();
|
|
for (auto& job : m_event_loop.m_pending_promises) {
|
|
// When this event loop was not running below another event loop, the jobs may very well have finished in the meantime.
|
|
if (!job->is_resolved())
|
|
job->cancel(Error::from_string_view("EventLoop is exiting"sv));
|
|
}
|
|
EventLoop::current().take_pending_events_from(m_event_loop);
|
|
}
|
|
}
|
|
|
|
private:
|
|
EventLoop& m_event_loop;
|
|
};
|
|
|
|
int EventLoop::exec()
|
|
{
|
|
EventLoopPusher pusher(*this);
|
|
for (;;) {
|
|
if (m_exit_requested)
|
|
return m_exit_code;
|
|
pump();
|
|
}
|
|
VERIFY_NOT_REACHED();
|
|
}
|
|
|
|
void EventLoop::spin_until(Function<bool()> goal_condition)
|
|
{
|
|
EventLoopPusher pusher(*this);
|
|
while (!goal_condition())
|
|
pump();
|
|
}
|
|
|
|
size_t EventLoop::pump(WaitMode mode)
|
|
{
|
|
wait_for_event(mode);
|
|
|
|
decltype(m_queued_events) events;
|
|
{
|
|
Threading::MutexLocker locker(m_private->lock);
|
|
events = move(m_queued_events);
|
|
}
|
|
|
|
m_pending_promises.remove_all_matching([](auto& job) { return job->is_resolved() || job->is_canceled(); });
|
|
|
|
size_t processed_events = 0;
|
|
for (size_t i = 0; i < events.size(); ++i) {
|
|
auto& queued_event = events.at(i);
|
|
auto receiver = queued_event.receiver.strong_ref();
|
|
auto& event = *queued_event.event;
|
|
if (receiver)
|
|
dbgln_if(EVENTLOOP_DEBUG, "Core::EventLoop: {} event {}", *receiver, event.type());
|
|
|
|
if (!receiver) {
|
|
switch (event.type()) {
|
|
case Event::Quit:
|
|
VERIFY_NOT_REACHED();
|
|
default:
|
|
dbgln_if(EVENTLOOP_DEBUG, "Event type {} with no receiver :(", event.type());
|
|
break;
|
|
}
|
|
} else if (event.type() == Event::Type::DeferredInvoke) {
|
|
dbgln_if(DEFERRED_INVOKE_DEBUG, "DeferredInvoke: receiver = {}", *receiver);
|
|
static_cast<DeferredInvocationEvent&>(event).m_invokee();
|
|
} else {
|
|
NonnullRefPtr<Object> protector(*receiver);
|
|
receiver->dispatch_event(event);
|
|
}
|
|
++processed_events;
|
|
|
|
if (m_exit_requested) {
|
|
Threading::MutexLocker locker(m_private->lock);
|
|
dbgln_if(EVENTLOOP_DEBUG, "Core::EventLoop: Exit requested. Rejigging {} events.", events.size() - i);
|
|
decltype(m_queued_events) new_event_queue;
|
|
new_event_queue.ensure_capacity(m_queued_events.size() + events.size());
|
|
for (++i; i < events.size(); ++i)
|
|
new_event_queue.unchecked_append(move(events[i]));
|
|
new_event_queue.extend(move(m_queued_events));
|
|
m_queued_events = move(new_event_queue);
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (m_pending_promises.size() > 30 && !s_warned_promise_count) {
|
|
s_warned_promise_count = true;
|
|
dbgln("EventLoop {:p} warning: Job queue wasn't designed for this load ({} promises). Please begin optimizing EventLoop::pump() -> m_pending_promises.remove_all_matching", this, m_pending_promises.size());
|
|
}
|
|
|
|
return processed_events;
|
|
}
|
|
|
|
void EventLoop::post_event(Object& receiver, NonnullOwnPtr<Event>&& event)
|
|
{
|
|
Threading::MutexLocker lock(m_private->lock);
|
|
dbgln_if(EVENTLOOP_DEBUG, "Core::EventLoop::post_event: ({}) << receiver={}, event={}", m_queued_events.size(), receiver, event);
|
|
m_queued_events.empend(receiver, move(event));
|
|
}
|
|
|
|
void EventLoop::add_job(NonnullRefPtr<Promise<NonnullRefPtr<Object>>> job_promise)
|
|
{
|
|
m_pending_promises.append(move(job_promise));
|
|
}
|
|
|
|
SignalHandlers::SignalHandlers(int signo, void (*handle_signal)(int))
|
|
: m_signo(signo)
|
|
, m_original_handler(signal(signo, handle_signal))
|
|
{
|
|
dbgln_if(EVENTLOOP_DEBUG, "Core::EventLoop: Registered handler for signal {}", m_signo);
|
|
}
|
|
|
|
SignalHandlers::~SignalHandlers()
|
|
{
|
|
dbgln_if(EVENTLOOP_DEBUG, "Core::EventLoop: Unregistering handler for signal {}", m_signo);
|
|
signal(m_signo, m_original_handler);
|
|
}
|
|
|
|
void SignalHandlers::dispatch()
|
|
{
|
|
TemporaryChange change(m_calling_handlers, true);
|
|
for (auto& handler : m_handlers)
|
|
handler.value(m_signo);
|
|
if (!m_handlers_pending.is_empty()) {
|
|
// Apply pending adds/removes
|
|
for (auto& handler : m_handlers_pending) {
|
|
if (handler.value) {
|
|
auto result = m_handlers.set(handler.key, move(handler.value));
|
|
VERIFY(result == AK::HashSetResult::InsertedNewEntry);
|
|
} else {
|
|
m_handlers.remove(handler.key);
|
|
}
|
|
}
|
|
m_handlers_pending.clear();
|
|
}
|
|
}
|
|
|
|
int SignalHandlers::add(Function<void(int)>&& handler)
|
|
{
|
|
int id = ++signals_info()->next_signal_id; // TODO: worry about wrapping and duplicates?
|
|
if (m_calling_handlers)
|
|
m_handlers_pending.set(id, move(handler));
|
|
else
|
|
m_handlers.set(id, move(handler));
|
|
return id;
|
|
}
|
|
|
|
bool SignalHandlers::remove(int handler_id)
|
|
{
|
|
VERIFY(handler_id != 0);
|
|
if (m_calling_handlers) {
|
|
auto it = m_handlers.find(handler_id);
|
|
if (it != m_handlers.end()) {
|
|
// Mark pending remove
|
|
m_handlers_pending.set(handler_id, {});
|
|
return true;
|
|
}
|
|
it = m_handlers_pending.find(handler_id);
|
|
if (it != m_handlers_pending.end()) {
|
|
if (!it->value)
|
|
return false; // already was marked as deleted
|
|
it->value = nullptr;
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
return m_handlers.remove(handler_id);
|
|
}
|
|
|
|
void EventLoop::dispatch_signal(int signo)
|
|
{
|
|
auto& info = *signals_info();
|
|
auto handlers = info.signal_handlers.find(signo);
|
|
if (handlers != info.signal_handlers.end()) {
|
|
// Make sure we bump the ref count while dispatching the handlers!
|
|
// This allows a handler to unregister/register while the handlers
|
|
// are being called!
|
|
auto handler = handlers->value;
|
|
dbgln_if(EVENTLOOP_DEBUG, "Core::EventLoop: dispatching signal {}", signo);
|
|
handler->dispatch();
|
|
}
|
|
}
|
|
|
|
void EventLoop::handle_signal(int signo)
|
|
{
|
|
VERIFY(signo != 0);
|
|
// We MUST check if the current pid still matches, because there
|
|
// is a window between fork() and exec() where a signal delivered
|
|
// to our fork could be inadvertently routed to the parent process!
|
|
if (getpid() == s_pid) {
|
|
int nwritten = write(s_wake_pipe_fds[1], &signo, sizeof(signo));
|
|
if (nwritten < 0) {
|
|
perror("EventLoop::register_signal: write");
|
|
VERIFY_NOT_REACHED();
|
|
}
|
|
} else {
|
|
// We're a fork who received a signal, reset s_pid
|
|
s_pid = 0;
|
|
}
|
|
}
|
|
|
|
int EventLoop::register_signal(int signo, Function<void(int)> handler)
|
|
{
|
|
VERIFY(signo != 0);
|
|
auto& info = *signals_info();
|
|
auto handlers = info.signal_handlers.find(signo);
|
|
if (handlers == info.signal_handlers.end()) {
|
|
auto signal_handlers = adopt_ref(*new SignalHandlers(signo, EventLoop::handle_signal));
|
|
auto handler_id = signal_handlers->add(move(handler));
|
|
info.signal_handlers.set(signo, move(signal_handlers));
|
|
return handler_id;
|
|
} else {
|
|
return handlers->value->add(move(handler));
|
|
}
|
|
}
|
|
|
|
void EventLoop::unregister_signal(int handler_id)
|
|
{
|
|
VERIFY(handler_id != 0);
|
|
int remove_signo = 0;
|
|
auto& info = *signals_info();
|
|
for (auto& h : info.signal_handlers) {
|
|
auto& handlers = *h.value;
|
|
if (handlers.remove(handler_id)) {
|
|
if (handlers.is_empty())
|
|
remove_signo = handlers.m_signo;
|
|
break;
|
|
}
|
|
}
|
|
if (remove_signo != 0)
|
|
info.signal_handlers.remove(remove_signo);
|
|
}
|
|
|
|
void EventLoop::notify_forked(ForkEvent event)
|
|
{
|
|
VERIFY_EVENT_LOOP_INITIALIZED();
|
|
switch (event) {
|
|
case ForkEvent::Child:
|
|
s_event_loop_stack->clear();
|
|
s_timers->clear();
|
|
s_notifiers->clear();
|
|
s_wake_pipe_initialized = false;
|
|
initialize_wake_pipes();
|
|
if (auto* info = signals_info<false>()) {
|
|
info->signal_handlers.clear();
|
|
info->next_signal_id = 0;
|
|
}
|
|
s_pid = 0;
|
|
return;
|
|
}
|
|
|
|
VERIFY_NOT_REACHED();
|
|
}
|
|
|
|
void EventLoop::wait_for_event(WaitMode mode)
|
|
{
|
|
fd_set rfds;
|
|
fd_set wfds;
|
|
retry:
|
|
|
|
// Set up the file descriptors for select().
|
|
// Basically, we translate high-level event information into low-level selectable file descriptors.
|
|
FD_ZERO(&rfds);
|
|
FD_ZERO(&wfds);
|
|
|
|
int max_fd = 0;
|
|
auto add_fd_to_set = [&max_fd](int fd, fd_set& set) {
|
|
FD_SET(fd, &set);
|
|
if (fd > max_fd)
|
|
max_fd = fd;
|
|
};
|
|
|
|
int max_fd_added = -1;
|
|
// The wake pipe informs us of POSIX signals as well as manual calls to wake()
|
|
add_fd_to_set(s_wake_pipe_fds[0], rfds);
|
|
max_fd = max(max_fd, max_fd_added);
|
|
|
|
for (auto& notifier : *s_notifiers) {
|
|
if (notifier->event_mask() & Notifier::Read)
|
|
add_fd_to_set(notifier->fd(), rfds);
|
|
if (notifier->event_mask() & Notifier::Write)
|
|
add_fd_to_set(notifier->fd(), wfds);
|
|
if (notifier->event_mask() & Notifier::Exceptional)
|
|
VERIFY_NOT_REACHED();
|
|
}
|
|
|
|
bool queued_events_is_empty;
|
|
{
|
|
Threading::MutexLocker locker(m_private->lock);
|
|
queued_events_is_empty = m_queued_events.is_empty();
|
|
}
|
|
|
|
// Figure out how long to wait at maximum.
|
|
// This mainly depends on the WaitMode and whether we have pending events, but also the next expiring timer.
|
|
Time now;
|
|
struct timeval timeout = { 0, 0 };
|
|
bool should_wait_forever = false;
|
|
if (mode == WaitMode::WaitForEvents && queued_events_is_empty) {
|
|
auto next_timer_expiration = get_next_timer_expiration();
|
|
if (next_timer_expiration.has_value()) {
|
|
now = Time::now_monotonic_coarse();
|
|
auto computed_timeout = next_timer_expiration.value() - now;
|
|
if (computed_timeout.is_negative())
|
|
computed_timeout = Time::zero();
|
|
timeout = computed_timeout.to_timeval();
|
|
} else {
|
|
should_wait_forever = true;
|
|
}
|
|
}
|
|
|
|
try_select_again:
|
|
// select() and wait for file system events, calls to wake(), POSIX signals, or timer expirations.
|
|
int marked_fd_count = select(max_fd + 1, &rfds, &wfds, nullptr, should_wait_forever ? nullptr : &timeout);
|
|
// Because POSIX, we might spuriously return from select() with EINTR; just select again.
|
|
if (marked_fd_count < 0) {
|
|
int saved_errno = errno;
|
|
if (saved_errno == EINTR) {
|
|
if (m_exit_requested)
|
|
return;
|
|
goto try_select_again;
|
|
}
|
|
dbgln("Core::EventLoop::wait_for_event: {} ({}: {})", marked_fd_count, saved_errno, strerror(saved_errno));
|
|
VERIFY_NOT_REACHED();
|
|
}
|
|
|
|
// We woke up due to a call to wake() or a POSIX signal.
|
|
// Handle signals and see whether we need to handle events as well.
|
|
if (FD_ISSET(s_wake_pipe_fds[0], &rfds)) {
|
|
int wake_events[8];
|
|
ssize_t nread;
|
|
// We might receive another signal while read()ing here. The signal will go to the handle_signal properly,
|
|
// but we get interrupted. Therefore, just retry while we were interrupted.
|
|
do {
|
|
errno = 0;
|
|
nread = read(s_wake_pipe_fds[0], wake_events, sizeof(wake_events));
|
|
if (nread == 0)
|
|
break;
|
|
} while (nread < 0 && errno == EINTR);
|
|
if (nread < 0) {
|
|
perror("Core::EventLoop::wait_for_event: read from wake pipe");
|
|
VERIFY_NOT_REACHED();
|
|
}
|
|
VERIFY(nread > 0);
|
|
bool wake_requested = false;
|
|
int event_count = nread / sizeof(wake_events[0]);
|
|
for (int i = 0; i < event_count; i++) {
|
|
if (wake_events[i] != 0)
|
|
dispatch_signal(wake_events[i]);
|
|
else
|
|
wake_requested = true;
|
|
}
|
|
|
|
if (!wake_requested && nread == sizeof(wake_events))
|
|
goto retry;
|
|
}
|
|
|
|
if (!s_timers->is_empty()) {
|
|
now = Time::now_monotonic_coarse();
|
|
}
|
|
|
|
// Handle expired timers.
|
|
for (auto& it : *s_timers) {
|
|
auto& timer = *it.value;
|
|
if (!timer.has_expired(now))
|
|
continue;
|
|
auto owner = timer.owner.strong_ref();
|
|
if (timer.fire_when_not_visible == TimerShouldFireWhenNotVisible::No
|
|
&& owner && !owner->is_visible_for_timer_purposes()) {
|
|
continue;
|
|
}
|
|
|
|
dbgln_if(EVENTLOOP_DEBUG, "Core::EventLoop: Timer {} has expired, sending Core::TimerEvent to {}", timer.timer_id, *owner);
|
|
|
|
if (owner)
|
|
post_event(*owner, make<TimerEvent>(timer.timer_id));
|
|
if (timer.should_reload) {
|
|
timer.reload(now);
|
|
} else {
|
|
// FIXME: Support removing expired timers that don't want to reload.
|
|
VERIFY_NOT_REACHED();
|
|
}
|
|
}
|
|
|
|
if (!marked_fd_count)
|
|
return;
|
|
|
|
// Handle file system notifiers by making them normal events.
|
|
for (auto& notifier : *s_notifiers) {
|
|
if (FD_ISSET(notifier->fd(), &rfds)) {
|
|
if (notifier->event_mask() & Notifier::Event::Read)
|
|
post_event(*notifier, make<NotifierReadEvent>(notifier->fd()));
|
|
}
|
|
if (FD_ISSET(notifier->fd(), &wfds)) {
|
|
if (notifier->event_mask() & Notifier::Event::Write)
|
|
post_event(*notifier, make<NotifierWriteEvent>(notifier->fd()));
|
|
}
|
|
}
|
|
}
|
|
|
|
bool EventLoopTimer::has_expired(Time const& now) const
|
|
{
|
|
return now > fire_time;
|
|
}
|
|
|
|
void EventLoopTimer::reload(Time const& now)
|
|
{
|
|
fire_time = now + interval;
|
|
}
|
|
|
|
Optional<Time> EventLoop::get_next_timer_expiration()
|
|
{
|
|
auto now = Time::now_monotonic_coarse();
|
|
Optional<Time> soonest {};
|
|
for (auto& it : *s_timers) {
|
|
auto& fire_time = it.value->fire_time;
|
|
auto owner = it.value->owner.strong_ref();
|
|
if (it.value->fire_when_not_visible == TimerShouldFireWhenNotVisible::No
|
|
&& owner && !owner->is_visible_for_timer_purposes()) {
|
|
continue;
|
|
}
|
|
// OPTIMIZATION: If we have a timer that needs to fire right away, we can stop looking here.
|
|
// FIXME: This whole operation could be O(1) with a better data structure.
|
|
if (fire_time < now)
|
|
return now;
|
|
if (!soonest.has_value() || fire_time < soonest.value())
|
|
soonest = fire_time;
|
|
}
|
|
return soonest;
|
|
}
|
|
|
|
int EventLoop::register_timer(Object& object, int milliseconds, bool should_reload, TimerShouldFireWhenNotVisible fire_when_not_visible)
|
|
{
|
|
VERIFY_EVENT_LOOP_INITIALIZED();
|
|
VERIFY(milliseconds >= 0);
|
|
auto timer = make<EventLoopTimer>();
|
|
timer->owner = object;
|
|
timer->interval = Time::from_milliseconds(milliseconds);
|
|
timer->reload(Time::now_monotonic_coarse());
|
|
timer->should_reload = should_reload;
|
|
timer->fire_when_not_visible = fire_when_not_visible;
|
|
int timer_id = s_id_allocator.with_locked([](auto& allocator) { return allocator->allocate(); });
|
|
timer->timer_id = timer_id;
|
|
s_timers->set(timer_id, move(timer));
|
|
return timer_id;
|
|
}
|
|
|
|
bool EventLoop::unregister_timer(int timer_id)
|
|
{
|
|
VERIFY_EVENT_LOOP_INITIALIZED();
|
|
s_id_allocator.with_locked([&](auto& allocator) { allocator->deallocate(timer_id); });
|
|
auto it = s_timers->find(timer_id);
|
|
if (it == s_timers->end())
|
|
return false;
|
|
s_timers->remove(it);
|
|
return true;
|
|
}
|
|
|
|
void EventLoop::register_notifier(Badge<Notifier>, Notifier& notifier)
|
|
{
|
|
VERIFY_EVENT_LOOP_INITIALIZED();
|
|
s_notifiers->set(¬ifier);
|
|
}
|
|
|
|
void EventLoop::unregister_notifier(Badge<Notifier>, Notifier& notifier)
|
|
{
|
|
VERIFY_EVENT_LOOP_INITIALIZED();
|
|
s_notifiers->remove(¬ifier);
|
|
}
|
|
|
|
void EventLoop::wake_current()
|
|
{
|
|
EventLoop::current().wake();
|
|
}
|
|
|
|
void EventLoop::wake()
|
|
{
|
|
dbgln_if(EVENTLOOP_DEBUG, "Core::EventLoop::wake()");
|
|
int wake_event = 0;
|
|
int nwritten = write((*m_wake_pipe_fds)[1], &wake_event, sizeof(wake_event));
|
|
if (nwritten < 0) {
|
|
perror("EventLoop::wake: write");
|
|
VERIFY_NOT_REACHED();
|
|
}
|
|
}
|
|
|
|
EventLoop::QueuedEvent::QueuedEvent(Object& receiver, NonnullOwnPtr<Event> event)
|
|
: receiver(receiver)
|
|
, event(move(event))
|
|
{
|
|
}
|
|
|
|
EventLoop::QueuedEvent::QueuedEvent(QueuedEvent&& other)
|
|
: receiver(other.receiver)
|
|
, event(move(other.event))
|
|
{
|
|
}
|
|
|
|
}
|