delayed write buffer question
I'm putting up a very basic question. Suppose we are using demand paging. We have a swap space partition on hard drive like we do in linux. The buffer policy is 'delayed-write'. Assume we are using global page replacement policy.
A process opens some files, and makes changes in them. The first time the page is brought into memory, it is taken from the filesystem (of course!).
Now, suppose another process wants an extra page, and the MMU page replacement algorithm chooses this particular page for replacement.
Now, will the OS just write the page to the swap space, or as the page is marked delayed-write, and it is being swapped, will the data be written back to the file-system area of the file also?
I know that if we are going to write it back to the file-system area, what speed advantage are we going to gain. But, just a question, and looking for a little elaborate answers. Basically I'm tring to really understand undo and redo, that's all.
Don't #$#$E@@% at me please.