[28896] in North American Network Operators' Group
Re: That pesky AS path corruption bug...
daemon@ATHENA.MIT.EDU (Peter T. Whiting)
Tue May 23 13:04:16 2000
Date: Tue, 23 May 2000 12:00:21 -0500
From: "Peter T. Whiting" <pwhiting@fury.ittc.ukans.edu>
To: Blaine Christian <blaine@inbound.blaines.net>
Cc: nanog@merit.edu
Message-ID: <20000523120021.A10571@sprint.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Pine.LNX.4.10.10005231234410.19459-100000@inbound.blaines.net>; from blaine@inbound.blaines.net on Tue, May 23, 2000 at 12:40:00PM -0400
Errors-To: owner-nanog-outgoing@merit.edu
When you say the router crashes I assume you are talking the whole box
crashing and not just the BGP session.
As I understand the current spec, a router, upon receiving a malformed
as_path is supposed to respond with a notification message (3.11) and
drop the BGP connection. Your suggestion to maintain the connection
and drop the announcement is a practical one, but doesn't put as much
pressure on vendors to fix the bug.
pete
On Tue, May 23, 2000 at 12:40:00PM -0400, Blaine Christian wrote:
>
> Hello all,
>
> After observing a recent issue regarding a router that sent corrupted AS
> path (all names are witheld to protect the guilty). I took a look at the
> path information that was being received and have a possible solution.
> Since the corrupted AS-path does not include the AS that the route is
> coming from (at least in the corruption that I saw) it seems to me that a
> simple solution for all is to filter on AS i.e. only allow routes that
> have the AS of your EBGP neighbor prepended to them. I realize this does
> not cover all cases of wacky AS corruption problems but it may fix some of
> them. I would suggest that those of you running mixed vendor EBGP (again
> names witheld) should implement a version of this strategy for your own
> self protection. It can certainly be implemented as part of an overall
> customer access functionality. This may be obvious to some of you but I
> do not believe that everyone is at this level yet.
>
> Of course the tirade part of this email is for all vendors involved in
> this travesty. If you do not understand or dislike a route that you have
> received don't just CRASH. Anyone ever thought of checking the route and
> throwing it out with an error message if you don't like it? I, of course,
> have heard and seen that several vendors have fixed this in the more
> recent releases. This type of bug is something that everyone who writes
> software has to deal with. If you raise an exception for bad input it is
> bad form to crash or reset your application.
>
> BTW, I am sure all have heard this argument before. I just wanted to get
> this topic renewed.
>
> Regards,
>
> Blaine
>
>
