Summary

Semaphores are specialized data structures used to coordinate access to a non-shareable resource (section of code). Cooperating (or possibly competing) processes use the semaphore(s) to determine if a specific resource is available. If the resource is unavailable, by default the system places the requesting process in an associated queue. The system notifies the waiting process when the resource is available. This alleviates the process from using polling to determine the availability of the resource. Semaphores can be categorized as binary or counting. Binary semaphores are used to synchronize access to a single instance of a non-shareable resource, while counting semaphores are used with multiple instances of a non-shareable resource.

The actions needed to manipulate semaphores are provided by a series of system calls. The semget system call is used to generate a new semaphore/ semaphore set (array) or to gain access to an existing semaphore. The semctl system call allows the user to set initial semaphore values, obtain their current value, and remove the semaphore. Operations on semaphores are performed with the semop system call. These operations (which are atomic) are used to decrement (obtain), increment (release), and test for zero-specific semaphores. Sets of operations can be specified if several semaphores are needed to coordinate access to a specific resource. The sets of operations may also be marked as being atomic.

While the syntax for using semaphores is somewhat complex, they do provide a standardized way of implementing classic primitive semaphore operations referenced in most operating system texts . As with many of the previous communication techniques, controlling access to a resource by using semaphores implies all involved processes will follow the rules. Semaphores cannot prevent processes from accessing a controlled resource.