On Thu, 10 Nov 2022 at 19:36, Vishal Moola <vishal.moola@xxxxxxxxx> wrote:
>
> On Tue, Nov 1, 2022 at 11:24 AM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
> >
> > On Tue, Nov 01, 2022 at 10:53:23AM -0700, Vishal Moola (Oracle) wrote:
> > > Converts the function to try to move folios instead of pages. Also
> > > converts fuse_check_page() to fuse_get_folio() since this is its only
> > > caller. This change removes 15 calls to compound_head().
> >
> > This all looks good. I wonder if we should't add an assertion that the
> > page we're trying to steal is !large? It seems to me that there are
> > assumptions in this part of fuse that it's only dealing with order-0
> > pages, and if someone gives it a page that's part of a large folio,
> > it's going to be messy. Miklos, any thoughts?
>
> Miklos, could you please look over this patch?
I think this should take care of the large folio case in fuse_try_move_page():
if (cs->len != PAGE_SIZE)
goto out_fallback;
The patch looks okay.
Acked-by: Miklos Szeredi <mszeredi@xxxxxxxxxx>
Thanks,
Miklos
