Hi,
realms are currently limited to 1 byte in the kernel, so it is probably
better to have some sort of 'tag to realm <start> [length]' with length
defaulting to 255 and something bigger than short - that way it will be
easy to make use of larger numbers for the realms later?
On 2016-07-01 16:16, Paul Jakma wrote:
On Fri, 1 Jul 2016, Piotr Chytla wrote:
Hi Paul,
Your route tags patches (forward ported by Cumulus) are queued up
for review for Quagga.
One question, should we perhaps use a typedef for the tag, for
internal purposes? Also, should it perhaps be wider than a
u_short?
My Route tags is part of bigger work , the idea was to push origin ASN
Number from bgpd to linux kernel as srcrealm ( in zebra as tag) and use
dstrealm as prefixlen. Realms were used by netflow probes fprobe /
softflowd
Everything is here http://packetconsulting.pl/~pch/securecon2007/
For me is ok if you want extend tag to longer type.
ASN and realms is exactly what I had in the back of my mind when I
asked about the width. :) It needs to be larger than a short these
days for that.
I'll mod the u_short to a typedef so.
Kaloyan - Cc'ed - is also interested in being able to set the realm on
routes sent to zebra. They had previously sent in patches to add an
explicit 'realm' internal attribute. However, it'd be better to keep
the routing protocol <-> zebra side generic with 'tags', and just map
'tags' to Linux 'realms' in the final step of zebra sending to the
kernel. Either with an option to just set the tag as realm, or else
some kind of more complex map of tag -> realm. So your 'tags' patch
should be a good basis for that, if you havn't got patches in that
area already.
regards,
_______________________________________________
Quagga-dev mailing list
[email protected]
https://lists.quagga.net/mailman/listinfo/quagga-dev
