Re: [RFC PATCH 5/5] geneve: add initial netdev driver for GENEVE tunnels

["John W. Linville" <linville@tuxdriver.com>]

On Mon, Apr 06, 2015 at 11:06:02AM -0700, Jesse Gross wrote:
> On Fri, Apr 3, 2015 at 6:01 PM, Francois Romieu <romieu@fr.zoreil.com> wrote:
> > Jesse Gross <jesse@nicira.com> :
> > [...]
> >> > diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c
> >> > new file mode 100644
> >> > index 000000000000..fe8895487fc2
> >> > --- /dev/null
> >> > +++ b/drivers/net/geneve.c
> >> > +/* geneve receive/decap routine */
> >> > +static void geneve_rx(struct geneve_sock *gs, struct sk_buff *skb)
> >> > +{
> >> > +       struct genevehdr *gnvh = geneve_hdr(skb);
> >> > +       struct geneve_dev *geneve;
> >> > +       struct pcpu_sw_netstats *stats;
> >> > +
> >> > +       geneve = gs->rcv_data;
> >> > +
> >> > +       /* Does the VNI match the device? */
> >> > +       if (memcmp(gnvh->vni, geneve->vni, sizeof(geneve->vni)))
> >> > +               goto drop;
> >>
> >> Since Geneve packets can carry options and this doesn't currently
> >> support any, I think we need to at least check the 'C' bit in the
> >> header and drop packets if it is set to ensure that we don't
> >> accidentally ignore critical options.
> >
> > Speaking of it, it's imho a bit too easy to confuse GENEVE_CRIT_OPT_TYPE
> > with the relevant 'C' bit mask.
> 
> Which 'C' bit mask? You mean the bitfield in the header? I guess but
> I'm not sure what would make it clearer and since they are different
> types it seems somewhat difficult to actually misuse them in practice.

What would you suggest, Francois?  A GENEVE_CRIT_OPT_PRESENT() macro?

John
-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.