Re: [PATCH v3 6/9] fuse: convert fuse_writepages_fill() to use a folio for its tmp page

[Josef Bacik <josef@toxicpanda.com>]

On Fri, Aug 23, 2024 at 02:38:02PM -0700, Joanne Koong wrote:
> On Fri, Aug 23, 2024 at 12:03 PM Josef Bacik <josef@toxicpanda.com> wrote:
> >
> > On Fri, Aug 23, 2024 at 09:27:27AM -0700, Joanne Koong wrote:
> > > To pave the way for refactoring out the shared logic in
> > > fuse_writepages_fill() and fuse_writepage_locked(), this change converts
> > > the temporary page in fuse_writepages_fill() to use the folio API.
> > >
> > > This is similar to the change in e0887e095a80 ("fuse: Convert
> > > fuse_writepage_locked to take a folio"), which converted the tmp page in
> > > fuse_writepage_locked() to use the folio API.
> > >
> > > inc_node_page_state() is intentionally preserved here instead of
> > > converting to node_stat_add_folio() since it is updating the stat of the
> > > underlying page and to better maintain API symmetry with
> > > dec_node_page_stat() in fuse_writepage_finish_stat().
> > >
> > > No functional changes added.
> > >
> > > Signed-off-by: Joanne Koong <joannelkoong@gmail.com>
> > > ---
> > >  fs/fuse/file.c | 14 +++++++-------
> > >  1 file changed, 7 insertions(+), 7 deletions(-)
> > >
> > > diff --git a/fs/fuse/file.c b/fs/fuse/file.c
> > > index a51b0b085616..905b202a7acd 100644
> > > --- a/fs/fuse/file.c
> > > +++ b/fs/fuse/file.c
> > > @@ -2260,7 +2260,7 @@ static int fuse_writepages_fill(struct folio *folio,
> > >       struct inode *inode = data->inode;
> > >       struct fuse_inode *fi = get_fuse_inode(inode);
> > >       struct fuse_conn *fc = get_fuse_conn(inode);
> > > -     struct page *tmp_page;
> > > +     struct folio *tmp_folio;
> > >       int err;
> > >
> > >       if (wpa && fuse_writepage_need_send(fc, &folio->page, ap, data)) {
> > > @@ -2269,8 +2269,8 @@ static int fuse_writepages_fill(struct folio *folio,
> > >       }
> > >
> > >       err = -ENOMEM;
> > > -     tmp_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
> > > -     if (!tmp_page)
> > > +     tmp_folio = folio_alloc(GFP_NOFS | __GFP_HIGHMEM, 0);
> > > +     if (!tmp_folio)
> > >               goto out_unlock;
> > >
> > >       /*
> > > @@ -2290,7 +2290,7 @@ static int fuse_writepages_fill(struct folio *folio,
> > >               err = -ENOMEM;
> > >               wpa = fuse_writepage_args_alloc();
> > >               if (!wpa) {
> > > -                     __free_page(tmp_page);
> > > +                     folio_put(tmp_folio);
> > >                       goto out_unlock;
> > >               }
> > >               fuse_writepage_add_to_bucket(fc, wpa);
> > > @@ -2308,14 +2308,14 @@ static int fuse_writepages_fill(struct folio *folio,
> > >       }
> > >       folio_start_writeback(folio);
> > >
> > > -     copy_highpage(tmp_page, &folio->page);
> > > -     ap->pages[ap->num_pages] = tmp_page;
> > > +     folio_copy(tmp_folio, folio);
> > > +     ap->pages[ap->num_pages] = &tmp_folio->page;
> > >       ap->descs[ap->num_pages].offset = 0;
> > >       ap->descs[ap->num_pages].length = PAGE_SIZE;
> > >       data->orig_pages[ap->num_pages] = &folio->page;
> > >
> > >       inc_wb_stat(&inode_to_bdi(inode)->wb, WB_WRITEBACK);
> > > -     inc_node_page_state(tmp_page, NR_WRITEBACK_TEMP);
> > > +     inc_node_page_state(&tmp_folio->page, NR_WRITEBACK_TEMP);
> >
> > I *think* you can use
> >
> > node_stat_add_folio(tmp_folio, NR_WRITEBACK_TEMP);
> >
> > here instead of inc_node_page_state().  Thanks,
> 
> I was thinking inc_node_page_state() here would be better for
> preserving the API symmetry with the dec_node_page_state() function
> that gets called when the writeback gets finished (in
> fuse_writepage_finish_stat) - I don't think it's immediately obvious
> that node_stat_add_folio() and dec_node_page_state() are inverses of
> each other. I don't feel strongly about this though, so i'm happy to
> change this to node_stat_add_folio as well.

Ah yeah that's a good point, probably better to convert those in one shot so
everything is consistent.  Thanks,

Josef