Re: [PATCH] media: videobuf2: Fix integer overrun in allocation

[Ricardo Ribalda]

Hi Laurent

On Wed, Mar 10, 2021 at 9:12 AM Laurent Pinchart
<laurent.pinchart@xxxxxxxxxxxxxxxx> wrote:
>
> Hi Ricardo,
>
> On Wed, Mar 10, 2021 at 08:58:39AM +0100, Ricardo Ribalda wrote:
> > On Wed, Mar 10, 2021 at 8:49 AM Laurent Pinchart wrote:
> > > On Wed, Mar 10, 2021 at 12:43:17AM +0100, Ricardo Ribalda wrote:
> > > > The plane_length is an unsigned integer. So, if we have a size of
> > > > 0xffffffff bytes we incorrectly allocate 0 bytes instead of 1 << 32.
> > > >
> > > > Cc: stable@xxxxxxxxxxxxxxx
> > > > Fixes: 7f8414594e47 ("[media] media: videobuf2: fix the length check for mmap")
> > > > Signed-off-by: Ricardo Ribalda <ribalda@xxxxxxxxxxxx>
> > > > ---
> > > >  drivers/media/common/videobuf2/videobuf2-core.c | 4 +++-
> > > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/media/common/videobuf2/videobuf2-core.c b/drivers/media/common/videobuf2/videobuf2-core.c
> > > > index 02281d13505f..543da515c761 100644
> > > > --- a/drivers/media/common/videobuf2/videobuf2-core.c
> > > > +++ b/drivers/media/common/videobuf2/videobuf2-core.c
> > > > @@ -223,8 +223,10 @@ static int __vb2_buf_mem_alloc(struct vb2_buffer *vb)
> > > >        * NOTE: mmapped areas should be page aligned
> > > >        */
> > > >       for (plane = 0; plane < vb->num_planes; ++plane) {
> > > > +             unsigned long size = vb->planes[plane].length;
> > >
> > > unsigned long is still 32-bit on 32-bit platforms.
> > >
> > > > +
> > > >               /* Memops alloc requires size to be page aligned. */
> > > > -             unsigned long size = PAGE_ALIGN(vb->planes[plane].length);
> > > > +             size = PAGE_ALIGN(size);
> > > >
> > > >               /* Did it wrap around? */
> > > >               if (size < vb->planes[plane].length)
> > >
> > > Doesn't this address the issue already ?
> >
> > Yes and no. If you need to allocate 0xffffffff you are still affected
> > by the underrun. The core will return an error instead of doing the
> > allocation.
> >
> > (yes, I know it is a lot of memory for a buffer)
>
> That's my point, I don't think there's a need for this :-) Especially
> with v4l2_buffer.m.offset being a __u32, we are limited to 4GB for *all*
> buffers.

I guess I will convert this patch into a documentation patch, so we
explicitly know the limit of the API
(1<<32 - PAGE_SIZE).

Thanks!

>
> --
> Regards,
>
> Laurent Pinchart



-- 
Ricardo Ribalda