What is an explanation of the term 'page fault'?

Answer 1

A page fault in any operating system will occur when a reference to memory indicates that the actual information has been paged out to disk, and is not available in memory at the present time.

When the page fault occurs, the system has to swap that page back into memory before it can reference the memory location. This is how virtual memory was implemented.

Page faults should be transparent to the program; it doesn't know or care if the memory was resident or not, since the net effect is to make that memory available to the program, whether it was paged out or not.

Answer 2

In modern operating systems, memory is allocated in discrete chunks, called pages. Whenever a program desired more memory, the operating system will provide it with space which is an integer multiple of the OS's page size. Pages can vary significantly depending on the hardware and OS settings, from 4 kilobytes to up to 2 gigabytes in size.

Modern systems use a concept called virtual memory, where the memory "space" doesn't automatically have to correspond to the amount of physical RAM in the system. Instead, the OS will allocate a memory page from this virtual space, and use whatever hardware is has to actually store it on. Thus, a memory page can be stored in RAM, or on flash disk, or a hard drive, or any other storage device designed for what is known as "backing store" (literally, what is "backing" the memory page).

Since only actual dynamic RAM is suitable for active manipulation, only pages which are in RAM can be used for work. When a process want to access a memory page that ISN'T stored in DRAM, but instead stored on hard drive (or flash, etc.), the OS generates a page fault. The page fault is a system call which tells the Virtual Memory subsystem of the OS to move the requested memory page from hard disk into DRAM (and, possibly move a page from DRAM onto hard drive).

So, in short, a page fault is a system call to move a memory page from hard drive into main memory.