> Jon Lewis reported a very interesting problem with gated &
> aliased interfaces . Ie: it gets really hard to eliminate the
> host routes from being broadcast. Jon uses ospf, although I found
> the same using rip .
eth0:* labels are not used more and gated works as it works in other OSes:
i.e. interfaces are identified only by their addresses.
What's about RIP: it controls exports by export {} statement, so that
you may filter these routes by normal access lists.
OSPF does not allow cheating by design. If an address exists,
it is advertised. This is controlled partially with "network" statement.
If you want to filter lots of aliases, you may add them also to not OSPFable
interfaces (f.e. dummy). But really creating bulk sets of aliases
if you do not want to export them is very bad idea.
Explicit aliases are created only to be used by gated, if it should
ignore them just make:
ip route add local 193.168.0.0/24 table local dev eth0 proto static
and linux will handle all these addresses as local with minimal cost.
(If the range does not form single prefixe, it may be split to several
subprefixes)
Alias labels eth0:xxx are not used anymore: they are preserved only
not to frighten too much our folks by BSdisms. You may use them
for convenience, when setting/deleting static routes,
but this information is not remembered by kernel and not used by gated.
> Also for another excersize just see what happens on a system where
> you are at the console ( I MEAN at the console ), Try creating
> the above aliased interface with a -really- simple gated.conf
> (see attached) and do a 'gdc stop' then a 'gdc start' .
> I think you'll be -VERY- surprised in the resultant route table .
I tried and not did not find any exciting effect.
What do you mean?
Alexey
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/