Re: Re: [PATCH v3] virtio-net: Add validation for used length

[Yongji Xie]

On Mon, May 31, 2021 at 2:49 PM Jason Wang <jasowang@xxxxxxxxxx> wrote:
>
>
> 在 2021/5/28 下午8:11, Xie Yongji 写道:
> > This adds validation for used length (might come
> > from an untrusted device) to avoid data corruption
> > or loss.
> >
> > Signed-off-by: Xie Yongji <xieyongji@xxxxxxxxxxxxx>
> > ---
> >   drivers/net/virtio_net.c | 28 +++++++++++++++++++++-------
> >   1 file changed, 21 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > index 073fec4c0df1..01f15b65824c 100644
> > --- a/drivers/net/virtio_net.c
> > +++ b/drivers/net/virtio_net.c
> > @@ -732,6 +732,17 @@ static struct sk_buff *receive_small(struct net_device *dev,
> >
> >       rcu_read_lock();
> >       xdp_prog = rcu_dereference(rq->xdp_prog);
> > +     if (unlikely(len > GOOD_PACKET_LEN)) {
> > +             pr_debug("%s: rx error: len %u exceeds max size %d\n",
> > +                      dev->name, len, GOOD_PACKET_LEN);
> > +             dev->stats.rx_length_errors++;
> > +             if (xdp_prog)
> > +                     goto err_xdp;
> > +
> > +             rcu_read_unlock();
> > +             put_page(page);
> > +             return NULL;
> > +     }
> >       if (xdp_prog) {
> >               struct virtio_net_hdr_mrg_rxbuf *hdr = buf + header_offset;
> >               struct xdp_frame *xdpf;
> > @@ -888,6 +899,16 @@ static struct sk_buff *receive_mergeable(struct net_device *dev,
> >
> >       rcu_read_lock();
> >       xdp_prog = rcu_dereference(rq->xdp_prog);
> > +     if (unlikely(len > truesize)) {
> > +             pr_debug("%s: rx error: len %u exceeds truesize %lu\n",
> > +                      dev->name, len, (unsigned long)ctx);
> > +             dev->stats.rx_length_errors++;
> > +             if (xdp_prog)
> > +                     goto err_xdp;
> > +
> > +             rcu_read_unlock();
> > +             goto err_skb;
> > +     }
>
>
> Patch looks correct but I'd rather not bother XDP here. It would be
> better if we just do the check before rcu_read_lock() and use err_skb
> directly() to avoid RCU/XDP stuffs.
>

If so, we will miss the statistics of xdp_drops. Is it OK?

Thanks,
Yongji