On Wed, Sep 24, 2025 at 12:47:22PM +1000, David Gibson wrote:
> On Tue, Sep 23, 2025 at 09:13:22PM -0400, Jon Maloy wrote:
[snip]
> > +			for (; RTA_OK(rta, na); rta = RTA_NEXT(rta, na)) {
> > +				switch (rta->rta_type) {
> > +				case NDA_DST:
> > +					dst = RTA_DATA(rta);
> > +					dstlen = RTA_PAYLOAD(rta);
> > +					break;
> > +				case NDA_LLADDR:
> > +					lladdr = RTA_DATA(rta);
> > +					lladdr_len = RTA_PAYLOAD(rta);
> > +					break;
> > +				default:
> > +					break;
> > +				}
> > +			}
> 
> It seems like somewhere there ought to be something specifying the
> address type of both DST and LLADDR, but I'm not sure exactly where.
> We should check those, just in case some weirdo runs us on a host with
> IPX over Token Ring.

Looked into this a bit more.  The LLADDR type is a property of the
link.  I don't know if we get an RTM_NEWLINK before RTM_NEWNEIGH on
the subscription channel.  If so, we can check that and give up if
it's not ARPHRD_ETHER.  If not, we should probably probe the LLADDR
type via our normal netlink socket and only subscribe if it is
ARPHRD_ETHER.

For the protocol address type, it looks to be in ndm->ndm_family,
though rtnetlink(7) doesn't make that very clear.  We should check
that, not just the address length.

-- 
David Gibson (he or they)	| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you, not the other way
				| around.
http://www.ozlabs.org/~dgibson