What exactly causes a page fault?

When a process is being executed not all the pages held by the process are loaded into the physical memory. Only those pages that are required are loaded. When the CPU runs out of memory it replaces the unused pages in the physical memory with the pages of the process that needs to be executed. Later, when the process whose page was removed from the physical memory tries to access the page, a page fault occurs during which the page is reloaded into the physical memory by removing some other page or utilizing any free memory that is available. Have a look at he following presentation slides for better understanding http://www.cs.ucla.edu/~ani/classes/cs111.08w/Notes/Lecture%2016.pdf

First of all let me explain how the mapping between logical address and physical address takes place. The physical address is divided into fixed size blocks called frames. Similarly the logical address is divided into blocks of same size called pages. Now the logical address generated by the cpu is divided into two parts page no and offset. A page table contains an index to the page no and corresponding mapping to the frame no. The offset is added to this frame no and that gives the corresponding address in the physical memory. The paging hardware is shown below : Now when a page no is not found in the page table a page fault occurs. So what happens when a page fault occurs .. 1. We check an internal table (usually kept with the process control block) for this process to determine whether the reference was a valid or an invalid memory access. 2. If the reference was invalid, we terminate the process. If it was valid, but we have not yet brought in that page, we now page it in. 3. We find a free frame (by taking one from the free-frame list, for example). 4. We schedule a disk operation to read the desired page into the newly allocated frame. 5. When the disk read is complete, we modify the internal table kept with the process and the page table to indicate that the page is now in memory. 6. We restart the instruction that was interrupted by the trap. The process can now access the page as though it had always been in memory. These steps can be understood by the image below  

When a process tries to visit a virtual address, the OS will first check the TLB, see if the page is cached there; if not, it refers to the page table, and check the valid bit is set or not; if not, then the mismatch arise, since the page is not in the frame. The OS has a global management to the mapping between virtual address space and physical space. Especially for a 64 bits system, since a program can be very large, and a multilevel paging scheme can be slow, the inverted page table will be adopted to map each page of each process to a frame (image this as a hash table, with separate chaining to solve collision). The mapping is NOT consistent, the OS will assign the free physical frame to processes which need more frames.