Re: [PATCH] mm/slab_common.c: fix possible spectre-v1 in kmalloc_slab()

[Dianzhang Chen]

thanks, i think your suggestion is ok.
in my previous method is easy to understand for spectre  logic,
but your suggestion is more sense to use of array_index_nospec.



On Thu, May 30, 2019 at 3:48 AM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
>
> On Wed, May 29, 2019 at 08:37:28PM +0800, Dianzhang Chen wrote:
> > The `size` in kmalloc_slab() is indirectly controlled by userspace via syscall: poll(defined in fs/select.c), hence leading to a potential exploitation of the Spectre variant 1 vulnerability.
> > The `size` can be controlled from: poll -> do_sys_poll -> kmalloc -> __kmalloc -> kmalloc_slab.
> >
> > Fix this by sanitizing `size` before using it to index size_index.
>
> I think it makes more sense to sanitize size in size_index_elem(),
> don't you?
>
>  static inline unsigned int size_index_elem(unsigned int bytes)
>  {
> -       return (bytes - 1) / 8;
> +       return array_index_nospec((bytes - 1) / 8, ARRAY_SIZE(size_index));
>  }
>
> (untested)