[lwip-devel] [task #13512] check how IPv6 handles the "no address" case for routing

[Simon Goldschmidt]

Follow-up Comment #5, task #13512 (project lwip):

> I'm not sure we should drop outgoing packets to an external network if we
have no routable address

Looking at the code, I think I need a bit of IPv6 lessons. As I see it, we
have two cases:
- sending to a link-local address or a subnet that matches one of the netifs,
which results pretty much in the same: sending a unicast ethernet packet to a
specific mac address
- sending to a non-link-local (i.e. routable) address that does not match any
local subnet. In this case we send to a router (much like IPv4 netif's 'gw').

Just to prevent misunderstandings: I wouldn't prevent sending to matching
subnets just because the netif doesn't have a routable address set. I would
prevent sending to non-local-match addresses on netifs without a routable
address. But this seems already the case in 'nd6_select_router()'.

As a result, there are 2 things to change left in ip6_route():
- when sending to local subnets, check up/link-up state of netifs
- when sending to non-local subnets, check if the netif returned by
nd6_select_router() is up/link-up.

However, in the 2nd case, it might be better to check up/link-up inside
nd6_select_router().

    _______________________________________________________

Reply to this item at:

  <http://savannah.nongnu.org/task/?13512>

_______________________________________________
  Nachricht gesendet von/durch Savannah
  http://savannah.nongnu.org/