illinois-cs341-system-programming/content/en/chap7/Semaphore.md

+++ title = "Semaphore" +++

Semaphore

A semaphore is another synchronization primitive. It is initialized to some value. Threads can either sem_wait or sem_post which lowers or increases the value. If the value reaches zero and a wait is called, the thread will be blocked until a post is called.

Using a semaphore is as easy as using a mutex. First, decide if on the initial value, for example the number of remaining spaces in an array. Unlike pthread mutex there are no shortcuts to creating a semaphore - use sem_init.

#include <semaphore.h>

sem_t s;
int main() {
  sem_init(&s, 0, 10); // returns -1 (=FAILED) on OS X
  sem_wait(&s); // Could do this 10 times without blocking
  sem_post(&s); // Announce that we've finished (and one more resource item is available; increment count)
  sem_destroy(&s); // release resources of the semaphore
}

When using a semaphore, wait and post can be called from different threads! Unlike a mutex, the increment and decrement can be from different threads.

This becomes especially useful if you want to use a semaphore to implement a mutex. A mutex is a semaphore that always waits before it posts. Some textbooks will refer to a mutex as a binary semaphore. You do have to be careful to never add more than one to a semaphore or otherwise your mutex abstraction breaks. That is usually why a mutex is used to implement a semaphore and vice versa.

• Initialize the semaphore with a count of one.

• Replace pthread_mutex_lock with sem_wait

• Replace pthread_mutex_unlock with sem_post

sem_t s;
sem_init(&s, 0, 1);

sem_wait(&s);
// Critical Section
sem_post(&s);

But be warned, it isn’t the same! A mutex can handle what we call lock inversion well. Meaning the following code breaks with a traditional mutex, but produces a race condition with threads.

// Thread 1
sem_wait(&s);
// Critical Section
sem_post(&s);

// Thread 2
// Some threads want to see the world burn
sem_post(&s);

// Thread 3
sem_wait(&s);
// Not thread-safe!
sem_post(&s);

If we replace it with mutex lock, it won’t work now.

// Thread 1
mutex_lock(&s);
// Critical Section
mutex_unlock(&s);

// Thread 2
// Foiled!
mutex_unlock(&s);

// Thread 3
mutex_lock(&s);
// Now it's thread-safe
mutex_unlock(&s);

Also, binary semaphores are different than mutexes because one thread can unlock a mutex from a different thread.

Signal Safety

Also, sem_post is one of a handful of functions that can be correctly used inside a signal handler pthread_mutex_unlock is not. We can release a waiting thread that can now make all of the calls that we disallowed to call inside the signal handler itself e.g. printf. Here is some code that utilizes this;

#include <stdio.h>
#include <pthread.h>
#include <signal.h>
#include <semaphore.h>
#include <unistd.h>

sem_t s;

void handler(int signal) {
  sem_post(&s); /* Release the Kraken! */
}

void *singsong(void *param) {
  sem_wait(&s);
  printf("Waiting until a signal releases...\n");
}

int main() {
  int ok = sem_init(&s, 0, 0 /* Initial value of zero*/);
  if (ok == -1) {
    perror("Could not create unnamed semaphore");
    return 1;
  }
  signal(SIGINT, handler); // Too simple! See Signals chapter

  pthread_t tid;
  pthread_create(&tid, NULL, singsong, NULL);
  pthread_exit(NULL); /* Process will exit when there are no more threads */
}

Other uses for semaphores are keeping track of empty spaces in arrays. We will discuss these in the thread-safe data structures section.