Page History

UNDER CONStRUCTION

Mutexes that are used to protect shared data structures being concurrently accessed. If a mutex is destroyed while a thread is blocked waiting for that mutex, critical sections (shared data that would otherwise be protected from data races) are no longer protected.

The C++ Standard, [thread.mutex.class], paragraph 5 [ISO/IEC 14882-2014], states:

The behavior of a program is undefined if it destroys a mutex object owned by any thread or a thread terminates while owning a mutex object.

accesses to shared data may be locked using the lock() member function and unlocked using the unlock() member function. If an exception occurs between the call to lock() and the call to unlock(), and the exception changes control flow such that unlock() is not called, the mutex will be left in the locked state and no critical sections protected by that mutex will be allowed to execute. This is likely to lead to deadlock.

The throwing of an exception must not allow a mutex to remain locked indefinitely. If a mutex was locked and an exception occurs within the critical section protected by that mutex, the mutex must be unlocked as part of exception handling before rethrowing the exception or continuing execution unless subsequent control flow will unlock the mutex.

C++ supplies the lock classes lock_guard, unique_lock, and shared_lock, which can be initialized with a mutex. In its constructor, the lock object locks the mutex, and in its destructor, it unlocks the mutex. The lock_guard class provides a simple RAII wrapper around a mutex. The unique_lock and shared_lock classes also use RAII and provide additional functionality, such as manual control over the locking strategy. The unique_lock class prevents the lock from being copied, although it allows the lock ownership to be moved to another lock. The shared_lock class allows the mutex to be shared by several locks. For all three classes, if an exception occurs and takes control flow out of the scope of the lock, the destructor will unlock the mutex and the program can continue working normally. These lock objects are the preferred way to ensure that a mutex is properly released when an exception is thrownThis statement implies that destroying a mutex while a thread is waiting on it is undefined behavior.

Noncompliant Code Example

This noncompliant code example creates several threads that each invoke the do_work() function, passing a unique number as an ID.

...

manipulates shared data and protects the critical section by locking the mutex. When it is finished, it unlocks the mutex. However, if an exception occurs while manipulating the shared data, the mutex will remain locked.

#include <mutex>
#include <thread>

constvoid sizemanipulate_t max_threads = 10;

void do_work(size_t i, shared_data(std::mutex *lockp&pm)
 {
  std::lock_guard<std::mutex> guard(*lockppm.lock();

  // AccessPerform datawork protectedon byshared the lock.
}

void start_threads(void)
{
  std::thread threads[max_threads];
  std::mutex lock;

  for (size_t i = 0; i < max_threads; ++i) {
    threads[i] = std::thread(do_work, i, &lock);
  }data.

  pm.unlock();
}

Compliant Solution (Manual Unlock)

This compliant solution eliminates the race condition by extending the lifetime of the lock:catches any exceptions thrown when performing work on the shared data and unlocks the mutex before rethrowing the exception.

#include <mutex>
#include <thread>

constvoid sizemanipulate_t max_threads = 10;

void do_work(size_t i, shared_data(std::mutex *lockp&pm)
 {
  std::lock_guard<std::mutex> guard(*lockppm.lock();
  try {
    // AccessPerform datawork protectedon byshared the lockdata.
}

std::mutex lock;

void start_threads(void)
{
  std::thread threads[max_threads];

  for (size_t i = 0; i < max_threads; ++i) {
    threads[i] = std::thread(do_work, i, &lock);
  }
}

Compliant Solution

  } catch (...) {
    pm.unlock();
    throw;
  }
  pm.unlock(); // in case no exceptions occur
}

Compliant Solution (Lock Object)

This compliant solution uses a lock_guard object to ensure that the mutex will be unlocked, even if an exception occurs, without relying on exception handling machinery and manual resource management.This compliant solution eliminates the race condition by joining the threads before the lock's destructor is invoked:

#include <mutex>
#include <thread>

constvoid sizemanipulate_t max_threads = 10;

void do_work(size_t i, shared_data(std::mutex *lockp&pm)
 {
  std::lock_guard<std::mutex> guardlk(*lockppm);

  // Access data protected by the lock.
}
void run_threads(void)
{
  std::thread threads[max_threads];
  std::mutex lock;

  for (size_t i = 0; i < max_threads; ++i) {
    threads[i] = std::thread(do_work, i, &lock);
  }

  for (size_t i = 0; i < max_threads; ++i) {
    threads[i].join();
  }
}

Risk Assessment

Perform work on shared data.
}

Risk Assessment

If an exception occurs while a mutex is locked, deadlock may resultDestroying a mutex while it is locked may result in invalid control flow and data corruption.

Automated Detection

Related Vulnerabilities

Search for vulnerabilities resulting from the violation of this rule on the CERT website.

Related Guidelines

This rule is a subset of ERR56-CPP. Guarantee exception safety.

Bibliography

...

Image Added Image Modified Image Removed  Image Modified