9.5: Mutex

My definition of Mutex is a wrapper for a type called pthread_mutex_t, which is defined in the POSIX threads API.

To create a POSIX mutex, you have to allocate space for a pthread_mutex_t type and then call pthread_mutex_init.

One of the problems with this API is that pthread_mutex_t behaves like a structure, so if you pass it as an argument, it makes a copy, which makes the mutex behave incorrectly. To avoid that, you have to pass pthread_mutex_t by address.

My code makes it easier to get that right. It defines a type, Mutex, which is just a more readable name for pthread_mutex_t:

#include <pthread.h>

typedef pthread_mutex_t Mutex;

Then it defines make_mutex, which allocates space and initializes the mutex:

Mutex *make_mutex()
{
    Mutex *mutex = check_malloc(sizeof(Mutex));
    int n = pthread_mutex_init(mutex, NULL);
    if (n != 0) perror_exit("make_lock failed"); 
    return mutex;
}

The return value is a pointer, which you can pass around as an argument without causing unwanted copying.

The functions to lock and unlock the mutex are simple wrappers for POSIX functions:

void mutex_lock(Mutex *mutex)
{
    int n = pthread_mutex_lock(mutex);
    if (n != 0) perror_exit("lock failed");
}

void mutex_unlock(Mutex *mutex)
{
    int n = pthread_mutex_unlock(mutex);
    if (n != 0) perror_exit("unlock failed");
}

This code is in mutex.c and the header file mutex.h.