What do the pthread_mutex_lock() and pthread_mutex_unlock() functions really do. I understand that the lock makes it so that the code is blocked until it is unlocked again. I'm still confused on what happens in between that period where it is locked and unlocked.

CodePudding user response：

In C, this is the basic way to lock and unlock mutexes is

    int pthread_mutex_lock(pthread_mutex_t *mutex);
    int pthread_mutex_unlock(pthread_mutex_t *mutex);

Basically pthread_mutex_unlock() will stop the lock and wait until the program asks for another lock to happen.

CodePudding user response：

At the most basic, every hardware out there has some sort of instruction or sequence of instructions that does "test and set": "As an atomic action, test if this variable has value x, and if so, set it to y. Let me know if I am successful".

So at it's heart, we have:

mutex_var = 0

pthread_mutex_lock(&mutex_var):
while True:
    if successful in atomically changing mutex_var from 0 to 1:
        break

pthread_mutex_unock(&mutex_var):
    mutex_var = 0

Now this code has two basic problems:

1. Anyone can unlock the mutex, even if they don't own it
2. This is wasteful of CPU. A thread waiting for the lock is in an infinite loop: "Can I have it?" "Can I have it?" "Can I have it?"

Actual implementations do a bit more. The mutex contains more information about who currently owns the mutex, if it is locked, so that only that thread can unlock it.

So the actual code is more like:

pthread_mutex_lock():
while True:
     if successful in atomically changing mutex_var from 0 to 1:
         break
     go to sleep waiting on this mutex

pthread_mutex_unlock():
     verify that this thread owns the mutex
     set mutex_var back to 0
     wake up all threads waiting on this mutex