Preserving device routes with kernel protocol

[Ondrej Zajicek]

On Tue, Apr 17, 2012 at 11:37:52AM +0100, Howden Nick wrote:
>    Hi,
> 
>    I tried to post this to the mailing list but to no avail.

Don't know why it wasn't delivered.

>    I wonder if you can shed some light on the problem we are having with
>    devices not being preserved when Bird updates the kernel routing tables.

>    >birdc show table BIRD
>    ...
>    11.6.9.0/24 via 11.6.1.200 on TAP_One [MYOSPF 16:04] * I (100/10436)
>    ...
>    The BIRD table contains the correct network device TAP_One

>    >ip route ls table NEW
>    ...
>    11.6.9.0/24 via 11.6.1.200 dev TAP_Two proto bird
>    ...
>    But, when we inspect the kernel table the network device information has
>    been lost.

Generally, BIRD assumed that IP address is the whole key for gw and
iface is derivable from it. With IPv6 and link-local addresses that had
to changed, but it probably remained in some parts where it is not
relevant to that. I personally never tested this variant.

>    Tap_One and Tap_Two both have IP addresses in the range 11.6.1.0/24 and I
>    am relying on the routing tables being correct in order to identify the
>    correct interface to use for the outbound traffic.  The behaviour we are
>    seeing is the behaviour we expect if routes are added as:
>    ip route add 11.6.9.0/24 via 11.6.1.200 table NEW
>    instead of
>    ip route add 11.6.9.0/24 via 11.6.1.200 dev TAP_One table NEW

You are right, for some unknown reason device is not used when route is
propagated to the kernel. You could try attached patch.

>    My understanding was that the "device routes" option would allow the
>    device part of the routes to be preserved but there is clearly a flaw in
>    my knowledge.

No, that 'device routes' is something completely different - whether
device routes (routes without via/gw) are propagated.

>    Is it actually possible to preserve the device information present in the
>    BIRD routing tables when they are exported to kernel tables?

Try attached patch.

-- 
Elen sila lumenn' omentielvo

Ondrej 'SanTiago' Zajicek (email: santiago at crfreenet.org)
OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net)
"To err is human -- to blame it on a computer is even more so."
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dev_in_routes.patch
Type: text/x-diff
Size: 843 bytes
Desc: not available
URL: <http://trubka.network.cz/pipermail/bird-users/attachments/20120417/b1f66098/attachment.patch>