Avoid deadlocks in the Alarm class by releasing the lock before invoking callbacks. This deadlock manifested itself in the ProgressManager: 1. On the main thread, the ProgressManager acquires its lock in ProgressManager::Decrement and calls Alarm::Create. 2. On the main thread, the Alarm acquires its lock in Alarm::Create. 3. On the alarm thread, the Alarm acquires its lock after waiting on the condition variable and calls ProgressManager::Expire. 4. On the alarm thread, the ProgressManager acquires its lock in ProgressManager::Expire. Note how the two threads are acquiring the locks in different orders. Deadlocks can be avoided by always acquiring locks in the same order, but since the two mutexes here are private implementation details, belong to different classes, that's not straightforward. Luckily, we don't need to have the Alarm mutex locked when invoking the callbacks. That exactly how this patch solves the issue.
223 lines
6.9 KiB
C++
223 lines
6.9 KiB
C++
//===-- Alarm.cpp ---------------------------------------------------------===//
|
|
//
|
|
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
|
|
// See https://llvm.org/LICENSE.txt for license information.
|
|
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#include "lldb/Host/Alarm.h"
|
|
#include "lldb/Host/ThreadLauncher.h"
|
|
#include "lldb/Utility/LLDBLog.h"
|
|
#include "lldb/Utility/Log.h"
|
|
|
|
using namespace lldb;
|
|
using namespace lldb_private;
|
|
|
|
Alarm::Alarm(Duration timeout, bool run_callback_on_exit)
|
|
: m_timeout(timeout), m_run_callbacks_on_exit(run_callback_on_exit) {
|
|
StartAlarmThread();
|
|
}
|
|
|
|
Alarm::~Alarm() { StopAlarmThread(); }
|
|
|
|
Alarm::Handle Alarm::Create(std::function<void()> callback) {
|
|
// Gracefully deal with the unlikely event that the alarm thread failed to
|
|
// launch.
|
|
if (!AlarmThreadRunning())
|
|
return INVALID_HANDLE;
|
|
|
|
// Compute the next expiration before we take the lock. This ensures that
|
|
// waiting on the lock doesn't eat into the timeout.
|
|
const TimePoint expiration = GetNextExpiration();
|
|
|
|
Handle handle = INVALID_HANDLE;
|
|
|
|
{
|
|
std::lock_guard<std::mutex> alarm_guard(m_alarm_mutex);
|
|
|
|
// Create a new unique entry and remember its handle.
|
|
m_entries.emplace_back(callback, expiration);
|
|
handle = m_entries.back().handle;
|
|
|
|
// Tell the alarm thread we need to recompute the next alarm.
|
|
m_recompute_next_alarm = true;
|
|
}
|
|
|
|
m_alarm_cv.notify_one();
|
|
return handle;
|
|
}
|
|
|
|
bool Alarm::Restart(Handle handle) {
|
|
// Gracefully deal with the unlikely event that the alarm thread failed to
|
|
// launch.
|
|
if (!AlarmThreadRunning())
|
|
return false;
|
|
|
|
// Compute the next expiration before we take the lock. This ensures that
|
|
// waiting on the lock doesn't eat into the timeout.
|
|
const TimePoint expiration = GetNextExpiration();
|
|
|
|
{
|
|
std::lock_guard<std::mutex> alarm_guard(m_alarm_mutex);
|
|
|
|
// Find the entry corresponding to the given handle.
|
|
const auto it =
|
|
std::find_if(m_entries.begin(), m_entries.end(),
|
|
[handle](Entry &entry) { return entry.handle == handle; });
|
|
if (it == m_entries.end())
|
|
return false;
|
|
|
|
// Update the expiration.
|
|
it->expiration = expiration;
|
|
|
|
// Tell the alarm thread we need to recompute the next alarm.
|
|
m_recompute_next_alarm = true;
|
|
}
|
|
|
|
m_alarm_cv.notify_one();
|
|
return true;
|
|
}
|
|
|
|
bool Alarm::Cancel(Handle handle) {
|
|
// Gracefully deal with the unlikely event that the alarm thread failed to
|
|
// launch.
|
|
if (!AlarmThreadRunning())
|
|
return false;
|
|
|
|
{
|
|
std::lock_guard<std::mutex> alarm_guard(m_alarm_mutex);
|
|
|
|
const auto it =
|
|
std::find_if(m_entries.begin(), m_entries.end(),
|
|
[handle](Entry &entry) { return entry.handle == handle; });
|
|
|
|
if (it == m_entries.end())
|
|
return false;
|
|
|
|
m_entries.erase(it);
|
|
}
|
|
|
|
// No need to notify the alarm thread. This only affects the alarm thread if
|
|
// we removed the entry that corresponds to the next alarm. If that's the
|
|
// case, the thread will wake up as scheduled, find no expired events, and
|
|
// recompute the next alarm time.
|
|
return true;
|
|
}
|
|
|
|
Alarm::Entry::Entry(Alarm::Callback callback, Alarm::TimePoint expiration)
|
|
: handle(Alarm::GetNextUniqueHandle()), callback(std::move(callback)),
|
|
expiration(std::move(expiration)) {}
|
|
|
|
void Alarm::StartAlarmThread() {
|
|
if (!m_alarm_thread.IsJoinable()) {
|
|
llvm::Expected<HostThread> alarm_thread = ThreadLauncher::LaunchThread(
|
|
"lldb.debugger.alarm-thread", [this] { return AlarmThread(); },
|
|
8 * 1024 * 1024); // Use larger 8MB stack for this thread
|
|
if (alarm_thread) {
|
|
m_alarm_thread = *alarm_thread;
|
|
} else {
|
|
LLDB_LOG_ERROR(GetLog(LLDBLog::Host), alarm_thread.takeError(),
|
|
"failed to launch host thread: {0}");
|
|
}
|
|
}
|
|
}
|
|
|
|
void Alarm::StopAlarmThread() {
|
|
if (m_alarm_thread.IsJoinable()) {
|
|
{
|
|
std::lock_guard<std::mutex> alarm_guard(m_alarm_mutex);
|
|
m_exit = true;
|
|
}
|
|
m_alarm_cv.notify_one();
|
|
m_alarm_thread.Join(nullptr);
|
|
}
|
|
}
|
|
|
|
bool Alarm::AlarmThreadRunning() { return m_alarm_thread.IsJoinable(); }
|
|
|
|
lldb::thread_result_t Alarm::AlarmThread() {
|
|
bool exit = false;
|
|
std::optional<TimePoint> next_alarm;
|
|
|
|
const auto predicate = [this] { return m_exit || m_recompute_next_alarm; };
|
|
|
|
while (!exit) {
|
|
// Synchronization between the main thread and the alarm thread using a
|
|
// mutex and condition variable. There are 2 reasons the thread can wake up:
|
|
//
|
|
// 1. The timeout for the next alarm expired.
|
|
//
|
|
// 2. The condition variable is notified that one of our shared variables
|
|
// (see predicate) was modified. Either the thread is asked to shut down
|
|
// or a new alarm came in and we need to recompute the next timeout.
|
|
//
|
|
// Below we only deal with the timeout expiring and fall through for dealing
|
|
// with the rest.
|
|
llvm::SmallVector<Callback, 1> callbacks;
|
|
{
|
|
std::unique_lock<std::mutex> alarm_lock(m_alarm_mutex);
|
|
if (next_alarm) {
|
|
if (!m_alarm_cv.wait_until(alarm_lock, *next_alarm, predicate)) {
|
|
// The timeout for the next alarm expired.
|
|
|
|
// Clear the next timeout to signal that we need to recompute the next
|
|
// timeout.
|
|
next_alarm.reset();
|
|
|
|
// Iterate over all the callbacks. Call the ones that have expired
|
|
// and remove them from the list.
|
|
const TimePoint now = std::chrono::system_clock::now();
|
|
auto it = m_entries.begin();
|
|
while (it != m_entries.end()) {
|
|
if (it->expiration <= now) {
|
|
callbacks.emplace_back(std::move(it->callback));
|
|
it = m_entries.erase(it);
|
|
} else {
|
|
it++;
|
|
}
|
|
}
|
|
}
|
|
} else {
|
|
m_alarm_cv.wait(alarm_lock, predicate);
|
|
}
|
|
|
|
// Fall through after waiting on the condition variable. At this point
|
|
// either the predicate is true or we woke up because an alarm expired.
|
|
|
|
// The alarm thread is shutting down.
|
|
if (m_exit) {
|
|
exit = true;
|
|
if (m_run_callbacks_on_exit) {
|
|
for (Entry &entry : m_entries)
|
|
callbacks.emplace_back(std::move(entry.callback));
|
|
}
|
|
}
|
|
|
|
// A new alarm was added or an alarm expired. Either way we need to
|
|
// recompute when this thread should wake up for the next alarm.
|
|
if (m_recompute_next_alarm || !next_alarm) {
|
|
for (Entry &entry : m_entries) {
|
|
if (!next_alarm || entry.expiration < *next_alarm)
|
|
next_alarm = entry.expiration;
|
|
}
|
|
m_recompute_next_alarm = false;
|
|
}
|
|
}
|
|
|
|
// Outside the lock, call the callbacks.
|
|
for (Callback &callback : callbacks)
|
|
callback();
|
|
}
|
|
return {};
|
|
}
|
|
|
|
Alarm::TimePoint Alarm::GetNextExpiration() const {
|
|
return std::chrono::system_clock::now() + m_timeout;
|
|
}
|
|
|
|
Alarm::Handle Alarm::GetNextUniqueHandle() {
|
|
static std::atomic<Handle> g_next_handle = 1;
|
|
return g_next_handle++;
|
|
}
|