OS X and iOS Kernel Programming

Locking

The use of atomic operations is a good solution for synchronizing access to a single variable. We often need to synchronize more complex sections of code, such as algorithms that rely on the value of multiple variables or functions that touch hardware. To remove the possibility of race conditions from these more complex areas of code, we rely on mutual exclusion; that is, any other thread that wishes to execute code which could interfere with the result of our operation is blocked until the operation is complete. The act of obtaining exclusive access is referred to as “acquiring a lock.”

The basic idea behind locking is that any code that accesses a shared resource, such as the instance variables of a driver, will first acquire a lock ...