Ok. Thanks for the clarification.

So if we look at softwire and NAT in isolation, the tunneling function
is fine with fragments. We need to do v4 defragmentation based on the
requirements of NAT function.

Defragmentation does open the avenue for "Fragmentation Buffer Full"
attack, but this will only affect subsequent fragments and not normal IP
packets. This is a tradeoff between supporting fragments to atleast an
extent.

Regards
Tarun Saxena

-----Original Message-----
From: Lee, Yiu [mailto:[email protected]] 
Sent: Wednesday, June 02, 2010 6:07 PM
To: Alain Durand
Cc: Tarun Saxena; [email protected]
Subject: Re: [Softwires] IPv4 fragmentation and reassembly in DS-Lite

Agreed. But the problem is the NAT device doesn't know how much buffer
it
will need to reserve for de-fragmentation because it doesn't know how
many
times a packet got fragmented. This opens hole for "Fragmentation Buffer
Full" attack. I am curious how today most home gateway handles v4
fragmentation from hosts?


On 6/2/10 8:11 AM, "Alain Durand" <[email protected]> wrote:

> Except that, as the AFTR needs to NAT the packet back to the tunnel,
it needs
> to reassemble the packet first..
> 
>    - Alain.
> 
> 
> 
_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to