On Tue, Feb 8, 2022 at 10:49 PM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
>
> On Tue, Feb 08, 2022 at 04:04:32PM +0800, Muchun Song wrote:
> > How about the name pvmw_set_pfn_from_page()?
> > When I looked at your next patch, I'm feeling this helper
> > is used to set up the pfn from a page instead of setting up
> > the page.
>
> But it doesn't just set up the pfn, it also sets up the range and
> the pgoff.
>
> I wonder if we wouldn't be better off with this pattern:
>
> DEFINE_PAGE_VMA_WALK(pvmw, page, vma, address, flags);
The approach sounds good to me. How about the following names?
DEFINE_PAGE_VMA_WALK_FROM_PAGE(pvmw, page, vma, address, flags);
This is used for most users that want to walk the whole page.
DEFINE_PAGE_VMA_WALK(pvmw, pfn, nr_pages, pgoff, vma, address, flags);
This is used for users that have no struct page or the user like
page_mapped_in_vma() which only wants to walk the part of
compound page.
Thanks.
>
> ... which wouldn't get used by KSM, but could be used by
> most other callers.
>
