SYSC3313 - Chapter 6 Part 3 Semaphores

Semaphore

Synchronization primitive based on a shared integer counter. P decrements when positive only and V increments.

Why use semaphores?

• can do mutual exclusion but no ownership principle

• task coordination

  • Signal between tasks

  • resource availability

Reader-Writers problem

Many tasks can access a shared resource at a time. Some read, some write. When writing, no other task can read or write.

Multiple readers allowed, but not multiple writers or both.

Readers-writers Solution

Use:

• Reader semaphore for readcount (init to 1)

• Writer semaphore for writing (init to 1)

• reader count = 0 initially

Readers:

1. Lock reader semaphore

2. Increment reader count

3. If first reader to enter, lock the writer semaphore

4. Unlock reader semaphore and read

5. Lock reader semaphore to update reader count

6. If last reader, unlock writer semaphore

7. Unlock reader semaphore

Writers:

1. Unlock writer semaphore

2. Write

3. Lock writer semaphore

Semaphore functions

sem_wait(sem_t *sem): Post/decrement operation

sem_post(sem_t * sem): Increment operation

sem_init(sem_t *sem, int pshared, unsigned int val): Init semaphore to value val. pshared = 0 means thread shareable versus process shareable

sem_destroy(sem_t *sem) deallocate semaphore

Create a named semaphore

To use a semaphore across independent executables (processes)

const char* SEM_NAME

sem_t *sem = sem_open(SEM_NAME, O_CREAT, 0666, 1)