Like other modern operating systems, Linux is multitasking, meaning that it can manage multiple running processes at the same time. With that great capability, though, comes great responsibility. Linux must ensure that no process can meddle with the memory used by another process. Historically, a vast number of security vulnerabilities were caused by malicious code being executed from a memory area that was intended for ordinary data storage only and not for executable code. The operating system also must ensure that enough memory is available for the needs of all running processes and must take steps to make memory available if there is not enough. All these responsibilities must be fulfilled as quickly as possible, or otherwise performance will suffer.

Fortunately, Linux provides a way to manage the memory resources for many disparate processes simultaneously: virtual memory [1]. Essentially, when a process refers to a memory address, it does not refer directly to a physical memory location. Instead, the memory address is used as an index into one or more tables, which are then used to translate the memory address into a real, physical memory address.

The concept of virtual memory is so well established that modern computer hardware almost always has at least some basic facilities built-in to make virtual memory management easier for the operating system. However, the exact details of how virtual memory is implemented varies from one hardware platform to another. Most of the concepts outlined in this article apply equally to any platform, but I have chosen to use the 64-bit x86 (PC) architecture as the basis for examples.

[...]