Chapter 8. Kernel Synchronization Introduction
In a shared memory application, care must be taken to ensure that shared resources are protected from concurrent access. The kernel is no exception. Shared resources require protection from concurrent access because if multiple threads of execution access and manipulate the data at the same time, the threads may overwrite each other's changes or access data while it is in an inconsistent state. Concurrent access of shared data is a recipe for instability that often proves very hard to track down and debuggetting it right off the bat is important.
Properly protecting shared resources can be tough. Years ago, before Linux supported symmetrical multiprocessing, preventing concurrent access of data was simple. Because only a single processor was supported, the only way data could have been accessed concurrently was if an interrupt occurred or if kernel code explicitly rescheduled and allowed another task to run. Back in those days, life was simple.
Those days are over. Symmetrical multiprocessing support was introduced in the 2.0 kernel and has been continually enhanced ever since. Multiprocessing support implies that kernel code can simultaneously run on two or more processors. Consequently, without protection, code in the kernel, running on two different processors, can simultaneously access shared data at exactly the same time. With the introduction of the 2.6 kernel, the Linux kernel is preemptive. This implies that (again, in the absence of protection) the scheduler can preempt kernel code at virtually any point and reschedule another task. Today, a number of scenarios allow for concurrency inside the kernel and they all require protection.
This chapter discusses the issues of concurrency and synchronization, as they exist in an operating system kernel. The next chapter details the mechanisms and interfaces that the Linux kernel provides to solve these synchronization issues and prevent race conditions.