What is the difference between Mutex and Binary semaphore?

First, "semaphore" is a fairly generic term. In common practice, there are 3
separate sets of UNIX-heritage interfaces for "semaphores". Two (System V
and POSIX) are "counting" semaphores, which have "values", and the other
(BSD) is a "binary" semaphore, which is either locked or not locked.

A mutex is essentially a binary semaphore, but with the additional property
that it is logically "owned" by the executable entity (whether process or
thread) that locked it. This property improves code structuring and allows
some useful optimizations. You can't lock a mutex in one thread and then
unlock it in another. You CAN, however, lock a binary semaphore in one
thread and unlock it in another, because while a binary semaphore may be
"locked" or "unlocked", it is never "owned".


A counting semaphore can be used to manage multiple resources, or for
generalized blocking rather than "locking". That is, a semaphore
initialized to the value 0 will always block when {waited, locked,
decreased}, and the waiter will be unblocked when it is {increased, posted,
unlocked}. If you initialize the semaphore to 1, the first "lock" will
succeed, while a second will block. If you have 5 shared buffers, you can
sometimes manage them simply by using a counting semaphore initialized to
the value 5. (The first 5 "locks" succeed, and subsequent "locks" block
until the semaphore is "unlocked".)

Senthil Kumar