Re: [PATCH 1/2] vdpa/mlx5: Fix umem sizes assignments on VQ create

["Michael S. Tsirkin" <mst@redhat.com>]

On Sun, May 30, 2021 at 11:26:46AM +0300, Eli Cohen wrote:
> On Sun, May 30, 2021 at 04:19:01AM -0400, Michael S. Tsirkin wrote:
> > On Sun, May 30, 2021 at 11:15:36AM +0300, Eli Cohen wrote:
> > > On Sun, May 30, 2021 at 04:05:16AM -0400, Michael S. Tsirkin wrote:
> > > > On Sun, May 30, 2021 at 09:31:28AM +0300, Eli Cohen wrote:
> > > > > Fix copy paste bug assigning umem1 size to umem2 and umem3.
> > > > > 
> > > > > Fixes: 1a86b377aa21 ("vdpa/mlx5: Add VDPA driver for supported mlx5 devices")
> > > > > Signed-off-by: Eli Cohen <elic@nvidia.com>
> > > > 
> > > > could you clarify the impact of the bug please?
> > > > 
> > > 
> > > I leads to firmware failure to create the virtqueue resource when you
> > > try to use a 1:1 mapping MR. This kind of usage is presented in the
> > > virtio_vdpa support I sent earlier.
> > 
> > OK pls include this info in the commit log.
> OK
> 
> > And is 1:1 the only case where
> > sizes differ? Is it true that in other cases sizes are all the same?
> > 
> The sizes are calculated based on firmware published paramters and a
> formula provided by the spec. They do differ but it so happened that
> size1 was larger than size2 and size3 so we did not see failures till
> now.

can this have a security impact? e.g. can guest access addresses
outside of it's memory with this?

> > > > > ---
> > > > >  drivers/vdpa/mlx5/net/mlx5_vnet.c | 4 ++--
> > > > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > > > 
> > > > > diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
> > > > > index 189e4385df40..53312f0460ad 100644
> > > > > --- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
> > > > > +++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
> > > > > @@ -828,9 +828,9 @@ static int create_virtqueue(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_virtque
> > > > >  	MLX5_SET(virtio_q, vq_ctx, umem_1_id, mvq->umem1.id);
> > > > >  	MLX5_SET(virtio_q, vq_ctx, umem_1_size, mvq->umem1.size);
> > > > >  	MLX5_SET(virtio_q, vq_ctx, umem_2_id, mvq->umem2.id);
> > > > > -	MLX5_SET(virtio_q, vq_ctx, umem_2_size, mvq->umem1.size);
> > > > > +	MLX5_SET(virtio_q, vq_ctx, umem_2_size, mvq->umem2.size);
> > > > >  	MLX5_SET(virtio_q, vq_ctx, umem_3_id, mvq->umem3.id);
> > > > > -	MLX5_SET(virtio_q, vq_ctx, umem_3_size, mvq->umem1.size);
> > > > > +	MLX5_SET(virtio_q, vq_ctx, umem_3_size, mvq->umem3.size);
> > > > >  	MLX5_SET(virtio_q, vq_ctx, pd, ndev->mvdev.res.pdn);
> > > > >  	if (MLX5_CAP_DEV_VDPA_EMULATION(ndev->mvdev.mdev, eth_frame_offload_type))
> > > > >  		MLX5_SET(virtio_q, vq_ctx, virtio_version_1_0, 1);
> > > > > -- 
> > > > > 2.31.1
> > > > 
> >