On Tue, Feb 26, 2002 at 06:25:34PM +0100, Maciej Soltysiak wrote:

> > They will?  Is that specific to 'icmp dst port'?  I thought routers
> > between the source and the destination could return ICMP errors with
> > their IP address if there is not route or such...
> Well, the normal scenario is this:
> A sends a packet to B to a closed port
> B sends imcp dest port unreach. The icmp dest port unreach has:
> source: B, dest: A, and as the payload 64 bytes (in case of linux) of the
> offending packets (there we see that A send that packet to B to a closed
> port)
> 
> The NEW scenario is this:
> A sends a packet to B to a closed port
> B sends icmp and changes the IP to a fake one.
> A receives the icmp, and says:
> hey, i did not send packets to 'fake IP', this is a mistake-->DROP
> 
> or is it not this way?
> 

Apparently not. I just tested it and it seems that A recieves the
ICMP port unreachable and parses the payload (note there are only
64 bits not bytes) and determines what the context was.

12:15:20.678887 eth0 > a.a.a.a.1031 > b.b.b.b.auth: S 1030380384:1030380384(0) ...
12:15:20.679323 eth0 < c.c.c.c > a.a.a.a: icmp: b.b.b.b tcp port auth unreachable (DF) 
[tos 0xc0] 

This is done by a simple '-j REJECT --reject-with icmp-port-unreachable'
on c.c.c.c. Now I assume that this new module would give you the
possibility to fake the originating icmp...

Ramin


Reply via email to