On Tue, Jun 11, 2024 at 11:21:31AM -0700, Andrew Morton wrote:
> On Tue, 11 Jun 2024 19:03:29 +0100 Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
>
> > On Tue, Jun 11, 2024 at 08:32:16AM -0700, Andrew Bresticker wrote:
> > > - if (page != &folio->page || folio_order(folio) != HPAGE_PMD_ORDER)
> > > + if (folio_order(folio) != HPAGE_PMD_ORDER)
> > > return ret;
> > > + page = &folio->page;
> >
> > This works today, but in about six months time it's going to be a pain.
> >
> > + page = folio_page(folio, 0);
> >
> > is the one which works today and in the future.
>
> I was wondering about that.
>
> hp2:/usr/src/25> fgrep "&folio->page" mm/*.c | wc -l
> 84
> hp2:/usr/src/25> fgrep "folio_page(" mm/*.c | wc -l
> 35
>
> Should these all be converted? What's the general rule here?
The rule is ...
- If we haven't thought about it, use &folio->page to indicate that
somebody needs to think about it.
- If the code needs to be modified to split folio and page apart, use
&folio->page.
- If the code is part of compat code which is going to have to be
removed, use &folio->page (eg do_read_cache_page()).
To *think* about it, and use folio_page() or folio_file_page(), don't
just blindly pass 0 as the second argument. Think about which page
within the folio is expected by the function you're working on.
Often that is "the first one!" and so folio_page(folio, 0) is the
right answer. But that should be justified.
It might be the right answer is "Oh, that function should take a folio".
