Let me make sure I understand C pthread mutex

Question

I have a global variable flag, a function this(), and a function that(). Now, main(), this() and that() all have occasions where they need to read and/or write flag. To properly use a mutex I would need to:

1. declare a pthread_mutex_t and initialize it.
2. create threads for this() and that() and pthread_join() them because I want main() to wait on this() and that() to finish before exiting.
3. anywhere in the code, be it main(), this(), or that(), where flag is being read or written, call pthread_mutex_lock() then [read/write] flag then call pthread_mutex_unlock()

Is that correct?

I don't plan on doing this but if I want to check flag in an infinite loop in main() like this:

    while(flag != value)
        continue;

Where would you lock() and unlock() the mutex and would the other threads get a chance to access flag with main() being constantly all over it like that?

Answer 1

1 - 2 - 3 - correct.

Now, if your write that loop, then

• as a consequence of 3, the main thread will hold the lock, and other threads won't be able to acquire the lock, so in effect, you created an infinite loop.
• if you don't lock, then it might not work, it depends on the CPU architecture (first of all, flag needs to be atomically updateable, and other cores need to see those changes. not all architectures guarantee this). Even if it works, it's still a busy wait loop, and shouldn't be used.

As Ben pointed out in his comment, you want to use condition variables for signaling (note: Pthread and wait conditions)

Where would you lock() and unlock() the mutex

As a thumb of rule, to minimize the contention, hold the lock as short as possible. This could be as simple as lock+read+unlock, but if the following sequence of operations depend on the fact that the flag shouldn't be changed, then keep the lock as long as needed.