Question

From here: https://computing.llnl.gov/tutorials/pthreads/#ConVarSignal

Note that the pthread_cond_wait
routine will automatically and atomically unlock mutex while it waits.

The following sub-code is from the same link (formatting by me):

pthread_mutex_lock(&count_mutex);

  while (count<COUNT_LIMIT) 
  {
      pthread_cond_wait(&count_threshold_cv, &count_mutex);
      printf("watch_count(): thread %ld Condition signal received.\n", my_id);
      count += 125;
      printf("watch_count(): thread %ld count now = %d.\n", my_id, count);
  }

pthread_mutex_unlock(&count_mutex);  

Question:
When it says that pthread_cond_wait will automatically unlock mutex while it waits, then why do we have to explicitly specify the function pthread_mutex_unlock at the end of the code above?

What’s the point that I am missing?

Answer 1

When pthread_cond_wait unblocks it is holding the lock again. Say for example you go around the loop twice you get the following sequence of lock/unlocks on the mutex:

lock

# Around loop twice:
    wait (unlock)
    awaken (holding lock)
    wait (unlock)
    awaken (holding lock)

# loop done, still holding lock

unlock

If you don’t have that last unlock there then you’ll end up with deadlock the next time someone else wants to acquire the lock.