Re: [PATCH] iomap: avoid avoid truncating 64-bit offset to 32 bits

[Marco Nelissen]

On Wed, Jan 8, 2025 at 8:38 PM Darrick J. Wong <djwong@xxxxxxxxxx> wrote:
>
> On Wed, Jan 08, 2025 at 08:11:50PM -0800, Marco Nelissen wrote:
> > on 32-bit kernels, iomap_write_delalloc_scan() was inadvertently using a
> > 32-bit position due to folio_next_index() returning an unsigned long.
> > This could lead to an infinite loop when writing to an xfs filesystem.
> >
> > Signed-off-by: Marco Nelissen <marco.nelissen@xxxxxxxxx>
> > ---
> >  fs/iomap/buffered-io.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> > index 54dc27d92781..d303e6c8900c 100644
> > --- a/fs/iomap/buffered-io.c
> > +++ b/fs/iomap/buffered-io.c
> > @@ -1138,7 +1138,7 @@ static void iomap_write_delalloc_scan(struct inode *inode,
> >                               start_byte, end_byte, iomap, punch);
> >
> >               /* move offset to start of next folio in range */
> > -             start_byte = folio_next_index(folio) << PAGE_SHIFT;
> > +             start_byte = folio_pos(folio) + folio_size(folio);
>
> eeek.  Yeah, I guess that would happen towards the upper end of the 16T
> range on 32-bit.

By "16T" do you mean 16 TeraByte? I'm able to reproduce the infinite loop
with files around 4 GB.

> I wonder if perhaps pagemap.h should have:
>
> static inline loff_t folio_next_pos(struct folio *folio)
> {
>         return folio_pos(folio) + folio_size(folio);
> }
>
> But I think this is the only place in the kernel that uses this
> construction?  So maybe not worth the fuss.
>
> Reviewed-by: "Darrick J. Wong" <djwong@xxxxxxxxxx>
>
> --D
>
> >               folio_unlock(folio);
> >               folio_put(folio);
> >       }
> > --
> > 2.39.5
> >