Re: [PATCH 5/6] openvswitch: Allow for any level of nesting in flow attributes

[Thomas Graf <tgraf@suug.ch>]

On 01/12/15 at 11:41am, Jesse Gross wrote:
> On Mon, Jan 12, 2015 at 4:26 AM, Thomas Graf <tgraf@suug.ch> wrote:
> > +       [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS]   = { .len = OVS_ATTR_NESTED },
> > +};
> 
> Geneve isn't really nested - maybe we should break it out into a
> separate name? OVS_ATTR_VARIABLE? We shouldn't really try to traverse
> it as netlink attributes anyways.

Agreed. I inteionally kept the behaviour intact in this series and was
about to do fix this in a separate patch to not make this series any
more complex than it already is ;-)

> > +static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
> > +       [OVS_KEY_ATTR_ENCAP]     = { .len = OVS_ATTR_NESTED },
> 
> This is not new but I think that encap isn't really handled correctly.
> In theory, there could be multiple levels of nesting here (either
> another encap or some other element) but there's no 'next' link.
> However, I don't believe the situation arises today.

Right. The behaviour is identical as before the patch. I just replaced
the -1 with a name. Agreed that we should clean this up