Re: [PATCH 10/12] vhost/crypto: fix build with GCC 12

[Bruce Richardson <bruce.richardson@intel.com>]

On Tue, Jun 14, 2022 at 11:22:24AM +0200, David Marchand wrote:
> On Thu, Jun 2, 2022 at 12:09 PM Bruce Richardson
> <bruce.richardson@intel.com> wrote:
> >
> > On Wed, May 18, 2022 at 12:16:55PM +0200, David Marchand wrote:
> > > GCC 12 raises the following warning:
> > >
> > > In file included from ../lib/mempool/rte_mempool.h:46,
> > >                  from ../lib/mbuf/rte_mbuf.h:38,
> > >                  from ../lib/vhost/vhost_crypto.c:7:
> > > ../lib/vhost/vhost_crypto.c: In function ‘rte_vhost_crypto_fetch_requests’:
> > > ../lib/eal/x86/include/rte_memcpy.h:371:9: warning: array subscript 1 is
> > >      outside array bounds of ‘struct virtio_crypto_op_data_req[1]’
> > >      [-Warray-bounds]
> > >   371 | rte_mov32((uint8_t *)dst + 3 * 32, (const uint8_t *)src + 3 * 32);
> > >       | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > ../lib/vhost/vhost_crypto.c:1178:42: note: while referencing ‘req’
> > >  1178 |         struct virtio_crypto_op_data_req req;
> > >       |                                          ^~~
> > >
> > > Check that copied length is within req boundaries.
> > >
> > > Fixes: 3c79609fda7c ("vhost/crypto: handle virtually non-contiguous buffers")
> > > Cc: stable@dpdk.org
> > >
> > > Signed-off-by: David Marchand <david.marchand@redhat.com>
> > > ---
> > >  lib/vhost/vhost_crypto.c | 8 ++++----
> > >  1 file changed, 4 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/lib/vhost/vhost_crypto.c b/lib/vhost/vhost_crypto.c
> > > index b1c0eb6a0f..83325b7042 100644
> > > --- a/lib/vhost/vhost_crypto.c
> > > +++ b/lib/vhost/vhost_crypto.c
> > > @@ -576,16 +576,16 @@ copy_data(void *dst_data, struct vhost_crypto_data_req *vc_req,
> > >       uint32_t to_copy;
> > >       uint8_t *data = dst_data;
> > >       uint8_t *src;
> > > -     int left = size;
> > > +     uint32_t left = size;
> > >
> > > -     to_copy = RTE_MIN(desc->len, (uint32_t)left);
> > > +     to_copy = RTE_MIN(desc->len, left);
> > >       dlen = to_copy;
> > >       src = IOVA_TO_VVA(uint8_t *, vc_req, desc->addr, &dlen,
> > >                       VHOST_ACCESS_RO);
> >
> > Tracking the functions which end up being called by this macro, the dlen
> > parameter ends up being of type "uint64_t *", passing a value of int * or
> > uint32_t * seems wrong to me. If we are changing the type from int to
> > uint32_t, I think it should be promoted all the way to uint64_t.
> 
> Indeed.
> I'll update in v2.
> 
> We already had some CVE on this part of the code, a careful review is needed.
> 
> 
> >
> > > -     if (unlikely(!src || !dlen))
> > > +     if (unlikely(!src || !dlen || dlen > left))
> > >               return -1;
> > >
> >
> > If this change is omitted, does the compiler still give warnings. Looking
> > through the called code, the dlen parameter can only ever be reduced, not
> > incremented (function rte_vhost_va_from_guest_pa() in rte_vhost.h).
> 
> If I promote to_copy and left variables as uint64_t, gcc is still
> unhappy, for the same reason.
> The check on dlen > left seems necessary.
> 
> 
Ok, just thought I'd ask anyway. I wonder if we need to check for
wrap-around in the reduction case, since we are dealing with unsigned
values. This additional check should catch that anyway if it does occur.

/Bruce