On Wed, 2004-06-30 at 20:05, IWAMOTO Toshihiro wrote:

At Wed, 30 Jun 2004 17:11:11 -0700,
Dave Hansen wrote:

quoted

On Wed, 2004-06-30 at 04:17, IWAMOTO Toshihiro wrote:

quoted

Due to struct page changes, page->mapping == NULL predicate can no
longer be used for detecting cancellation of an anonymous page
remapping operation.  So the PG_again bit is being used again.
It may be still possible to kill the PG_again bit, but the priority is
rather low.

But, you reintroduced it everywhere, including file-backed pages, not
just for anonymous pages?  Why was this necessary?

Which PG_again check are you talking about?
I think BUG_ON()s in file backed page codes should be kept for now.

I'm referring to all of the code segments like this:
        
        +       if (PageAgain(page)) {
        +               unlock_page(page);
        +               page_cache_release(page);
        +               goto again;
        +       }
        +       BUG_ON(PageAgain(page));
        
For any page that's in the page cache that you want to remove, simply
take the page lock, overwrite page->mapping, and the page cache code
will take care of the rest, no PG_again needed.  

For swap pages, one possibility to reserve a special swap entry
constant (SWAP_AGAIN) and check page->private instead of PageAgain
check, but I'm not sure if this is a good idea.

#define	SWAP_AGAIN	~0UL

Do you mean pages in the swap cache?

-- Dave

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>