std::timed_mutex::try_lock* fail spuriously

Question

By try_lock*, I take to mean try_lock(), try_lock_for(), and try_lock_until(). According to cppreference, all three methods may just fail spuriously. Following is quoted from the description for try_lock_for()

As with try_lock(), this function is allowed to fail spuriously and return false even if the mutex was not locked by any other thread at some point during timeout_duration.

I know that spurious wakeup may happen with std::condition_variable and the rationale behind it. But, what is the case with a mutex?

Answer 1

From C++14 chapter "30.4.1.2 Mutex types"

paragraph 16:

An implementation may fail to obtain the lock even if it is not held by any other thread. [Note: This spurious failure is normally uncommon, but allows interesting implementations based on a simple compare and exchange (Clause 29). —end note] An implementation should ensure that try_lock() does not consistently return false in the absence of contending mutex acquisitions.

and paragraph 19:

little would be known about the state after a failure, even in the absence of spurious failures

And in answer to

I know that spurious wakeup may happen with std::condition_variable and the rationale behind it. But, what is the case with a mutex?

std::timed_mutex sometimes is implemented using std::condition_varible when there is no direct support in the OS. As in GNU libstdc++:

#if _GTHREAD_USE_MUTEX_TIMEDLOCK

...

#else // !_GTHREAD_USE_MUTEX_TIMEDLOCK

  class timed_mutex
  {
    mutex       _M_mut;
    condition_variable  _M_cv;
    bool        _M_locked = false;

  public:

    template<typename _Rep, typename _Period>
      bool
      try_lock_for(const chrono::duration<_Rep, _Period>& __rtime)
      {
        unique_lock<mutex> __lk(_M_mut);
        if (!_M_cv.wait_for(__lk, __rtime, [&]{ return !_M_locked; }))
          return false;
        _M_locked = true;
        return true;
      }

    template<typename _Clock, typename _Duration>
      bool
      try_lock_until(const chrono::time_point<_Clock, _Duration>& __atime)
      {
        unique_lock<mutex> __lk(_M_mut);
        if (!_M_cv.wait_until(__lk, __atime, [&]{ return !_M_locked; }))
          return false;
        _M_locked = true;
        return true;
      }
  };

#endif

Answer 2

According to: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3209.htm

On the other hand, there are strong reasons to require that programs be written to tolerate spurious try_lock() failures:

1. As pointed out in Boehm, Adve, "Foundations of the C++ Concurrency Memory Model", PLDI 08, enforcing sequential consistency for data-race-free programs without spurious try_lock() failures requires significantly stronger memory ordering for lock() operations on try_lock()-compatible mutex types. On some architectures that significantly increases the cost of uncontended mutex acquisitions. This cost appears to greatly outweigh any benefit from prohibiting spurious try_lock() failures.
2. It allows a user-written try_lock() to fail if, for example, the implementation fails to acquire a low-level lock used to protect the mutex data structure. Or it allows such an operation to be written directly in terms of compare_exchange_weak.
3. It ensures that client code remains correct when, for example, a debugging thread is introduced that occasionally acquires locks in order to be able to read consistent values from a data structure being checked or examined. Any code that obtains information from try_lock() failure would break with the introduction of another thread that purely locks and reads the data structure.