Ahh thanks, missed that.

Get a better book

Programs are generally not moved to swap.

Most of a program is typically mapped into the process's address space from the original file. It is paged in from the original file as needed. If memory is tight, program pages that have not been used recently are dropped from memory (not written to swap). When such pages are needed again, they are paged in again from the original file.

The common, but wrong, idea you quoted is the root of a lot of common misconceptions about swap space. Swap space is used almost exclusively for paging (to from disk) "anonymous" pages. If you have too little swap space or restrict its use with swappiness, that causes more total paging, because program content (and other non anonymous pages) are often paged many times for lack of the ability to page anonymous data few times.

Hehe
So to use a link from earlier in this post, https://help.ubuntu.com/community/SwapFaq
"Swap space temporarily holds memory pages that are inactive."
You say not been used recently, I would assume you mean inactive as well ?

The important point I was making is that swap space is not involved at all when dropping pages from memory when those pages were file mapped (which is the usual method for most program content).


Which simplification you choose to describe the basis for selecting pages to drop, is a separate question. The reality is more complicated than the description. But "inactive" or "not used recently" are both reasonable as simple descriptions for a more complicated process.

By contrast I think "Swap space temporarily holds" is not a valid simple description for pages that never go to swap space.

I understand, thanks.