Linux 3.9 routing oddity

[Hannes Frederic Sowa]

Hello!

(I am the author of the patch discussed here.)

On Thu, Jul 04, 2013 at 02:40:50PM +0200, Sander Steffann wrote:
> > "If we don't compile with CONFIG_IPV6_ROUTER_PREF (RFC4191 support) a
> > neighbour is only valid if its nud_state is NUD_VALID. I did not find
> > any references that we should probe the router on route insertion time
> > via the other RFCs. So skip this route in that case."
> 
> "We didn't find an RFC that tells us to build a working implementation, so we decided not to..."

Of coure I did notice the misbehaviour if we compile without
CONFIG_IPV6_ROUTER_PREF and am working on a solution. I hoped to send
out a second patch soon after the first one, but it became more subtle
(and meanwhile lack time until the end of the weekend :( ).

This is a more complex issue to fix as I did not want to break the
round-robin selection of routes with unkown nud states in the default
router list. Also I am looking to resolve some usability issues: on
executing ``ip route get'' we actually do trigger active nud validation
on the nexthops. This could be a PITA because we do actually change the
results by that (a second ``ip route get'' could yield different results).

> > So I assume that the bug would still be there.
> 
> Sigh...

I'll keep you notified as soon as I have a working solution.

Greetings,

  Hannes