A solution to avoid race conditions is to use a tool offered by most operating systems: semaphores. A semaphore is a special integer variable that is used for controlling access to shared resources. The PThreadlibrary defines a Semaphore object in semaphore.h. There are 3 types of operations defined on a semaphore:

Initialization: In C, this is done by calling the sem_init function. For example the following initializes a shared semaphore (called mutex) to 1.

sem_t mutex = sem_init(&mutex, 0, 1);

Decrement and Wait: The value of the semaphore is decremented and if its value becomes negative, then the thread is blocked. This can be done by calling:

sem_wait(&mutex);

Increment and Signal: The value of the semaphore is incremented and if its value becomes negative or equal to zero, then a thread that was blocked is unblocked.In Java, this can be done by calling:

sem_post(&mutex);

Binary semaphores are simply semaphores that can take on the values 0 and 1.

Modify the file rc.c:

#include

#include

#include //compile and link with -pthread

#include

int global = 0;

sem_t mutex;

void *inc( void *ptr ) {

int i;

for(i = 0; i < 9000000; i++){

sem_wait(&mutex);

global++;

sem_post(&mutex);

}

}

void *dec( void *ptr ) {

int i;

for(i = 0; i < 9000000; i++){

sem_wait(&mutex);

global--;

sem_post(&mutex);

}

}

int main() {

pthread_t thread1, thread2;

sem_init(&mutex, 0, 1);

pthread_create( &thread1, NULL, inc, NULL);

pthread_create( &thread2, NULL, dec, NULL);

pthread_join( thread1, NULL);

pthread_join( thread2, NULL);

sem_destroy(&mutex);

printf("Final global = %d ", global);

exit(0);

}

Compile the program again:

student@Ubuntu:~/labs/lab3$ gcc -pthread rc.c -o rc

Run the program several times again:

student@Ubuntu:~ /labs/lab3 $ ./rc

What are the possible outputs this time? Explain in detail how the use of semaphores guarantees that the critical sections:

global++

global--

wont be accessed concurrently by the two threads.