capstone blog

Previously in the posts, GDT was the topic and the Wikipedia article with the code snippet was basically setting up kernel code segment and kernel data segment for paging. Remember the granularity flag, this was set to units of paging or in other words, set to 1. Let’s get into paging for this blog. It is highly relevant to the capstone, as the team is implementing a paging scheme within the final product, the Minimalistic OS.

As a reminder, both segment and paging are types of memory management methods. The virtual memory is actually a label to call a memory process of which the RAM/physical/main memory works together with a secondary storage device, such as a hard disk. More details about this virtual memory — later!

Content Sources

• Wikipedia
• wiki.osdev.org/Paging

Paging, like pages of a book?

Sort of? Paging is a way to make the individual processes running feel like they are the only ones running on the disk. Much like a book, the reader is fixed on all the information contained on one page, then continued to the next page, however still solely focused on the page at hand. Another book related perspective could be that the author has their whole masterpiece on a continuous paper spool. To make a book, the spool is divided into equal sized chunks (pages) and stitched together to make a book.

With paging, a very nice and convenient sequenced logical memory space is provided for each of the programs. This is usually 4KB chunks or pages. However, this is all an illusion, as these individual processes are not actually alone, they are joined by others who also take part into this grand illusion, known as paging. Continuous memory allocation is what the individual processes will be using within the logical memory space probably is not continuous (or at least has the option of not having to be) within the physical memory space, in fact, this was a main driver for the creation of paging memory management.

Paging will require a type of look-up table or what is known as the page table, which holds all the pages. This table will be pointed to from a paging directory entry. In this table the logical address (the one the illusion-ed individual process will be using) will be mapped to an actual, no illusion, physical memory address or if using virtual memory scheme then, there is a possibility the mapping will be to a special place within a hard disk or the like. However, even if using a virtual memory scheme or not, the page table will map the illusion logical memory to a physical place in memory.

Let us first start off with some chunk located in the logical address space. This logical address will have two pieces to its address: selector and offset. The selector will be used to point to something handy in the descriptor table, of which will be the particular segment descriptor. The offset will be used with the segment descriptor to create the linear address. The linear address contains the page directory entry, page table entry, and the offset within the physical address hosting the page frame.

Virtual Memory Purpose

Virtual memory was originally thought to help out the limitations of a RAM/main memory size. As stated before, the virtual memory scheme is a usage between the RAM/main memory and a secondary storage device, such as a hard drive. The RAM/main memory will accumulate memory chunks (pages), with an allocated size, all the same for paging granularity, then begin moving the unused memory chunks (pages) into a designated storage place within this secondary memory device. In a sense, the virtual memory scheme creates another illusion of a huge and ample space “main memory”. However, in reality, this actually is far from the case.

We have all these unused pages/memory chunks within the secondary storage device waiting, once they are needed, they are placed back to the RAM/main memory place. The page table is consistently working to keep track of all these memory chunks. It can be a daunting task, surely!

Troublesome Paging

Disk thrashing and fragmented processes are the main issues with the use of paging. If you were to use segmentation memory management these are not issues, however with segmentation you are not using every last available memory space and memory fragmentation effects will be seen soon. Trade offs.

Disk thrashing limits application-level processing by page faults and over worked virtual memory abilities. It is data swapping at its limits! The moving of data does not allow for effective storage in the physical memory.

fragmented processes are processes which really don’t have all that they need to carry out their purpose. This becomes present when the hosting of the data paging is not adequate enough to piece all together during the use of a process from physical memory.

Best to decided what is the goal of your system — utilize main memory effectively (paging) or aim for fast access by processor (segmentation). Modern systems tend towards paging and a hybrid of the two types.

Thank you for reading.

Tags: memory management, operating system, Paging, virtual memory