Truthfully, after spending 5 hours with a cell phone to my head I couldn't
be bothered asking how a router would be aware of a double NAT.
The problem only occurred when an ACK was returning with a destination that
had to be reached by the default route. The correct sequence of NAT
requires the NAT process to check for a route to the destination before
performing the translation. If there's a destination in the route table,
translation is supposed to occur. Once examined, it was clear that a
packet entering the network was Natted once on the 1720 and again on the
PIX, finally reaching the destination machine. When the destination
machine responded with an ACK, the PIX was translating correctly, but the
1720 was only translating if there was a specific entry in the route table
for the destination network. If the destination required the packet to
take the default route, no translation would occur and the packet was
dropped. The workaround required policy routing that looks at the source
of the ACK and then sets the "ip next-hop" manually if the packet matches
the rules.
Craig
At 07:44 AM 5/25/2001 -0700, you wrote:
>I agree with the other guy. How would any router "know" that a packet had
>already been natted?
>
>Does Cisco NAT set on of the IP or TCP obscure bits? What did Cisco say?
>
>Chuck
>
>-----Original Message-----
>From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] On Behalf Of
>Craig Columbus
>Sent: Friday, May 25, 2001 5:03 AM
>To: [EMAIL PROTECTED]
>Subject: Re: Double-Nat troubles [7:5752]
>
>Thanks for the reply Jason.
>
>Actually, the routing was completely correct. I spent 5 hours on the phone
>with TAC yesterday and they confirmed that it was a bug with the double
>NAT. For some reason, IOS NAT, when presented with already NATed traffic,
>won't properly send the traffic to a default route. Strangely enough, I
>was able to test another double-NAT situation, this time with two PIX
>boxes, and I had no issues at all. As a workaround, I had to put policy
>routing in place with a bunch of statements for the servers that needed to
>"ACK" traffic from the old network.
>
>Thanks again,
>Craig
>
>At 03:24 AM 5/25/2001 -0400, you wrote:
> >What if you move the default route to toward the PIX? I bet it works
then.
> >How is the router to know where to forward the packets just because there
>is
> >a NAT in place? The NAT happens as the packets go from one interace to
the
> >other, but that is still depending on routing taking place for it to know
> >where to send it. Why not just set up a static route toward the PIX for
>the
> >old network?
> >
> >--
> >Jason Roysdon, CCNP+Security/CCDP, MCSE, CNA, Network+, A+
> >List email: [EMAIL PROTECTED]
> >Homepage: http://jason.artoo.net/
> >
> >
> >
> >""Craig Columbus"" wrote in message
> >[EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> > > I'm about to open a case with Cisco regarding this issue, but I'm
>curious
> > > if anyone out there has run into something similar:
> > >
> > > Background:
> > > A company is migrating networks. For various reasons, the following
> > > network is in place.
> > > Cisco 1720: has a fast ethernet connecting to the PIX, an ethernet
> > > connecting to the old network, and a Serial connecting to the Internet.
> > > Ethernet0 Interface has address 1.1.1.3/24, routing all "old-network"
> > > traffic to 1.1.1.1. This is also a NAT Outside interface.
> > > Fast Ethernet0 has address 2.2.2.1/26, routing all inbound "new
network"
> > > traffic to 2.2.2.2 (PIX Outside Interface). This is a NAT Inside
> >interface.
> > > Serial 0 is the default-route for all destinations not found in the
>route
> > > table.
> > >
> > > Problem:
> > > When traffic is sent to a 1.1.1.x address, it gets translated properly
> >into
> > > a 2.2.2.x address and routed to the PIX. The PIX translates the
2.2.2.x
> > > address to a (static) 10.x.x.x address. The traffic reaches the
> > > destination machine inside the network. The machine then tries to
>respond
> > > to the original source. The PIX recognizes and properly translates the
> > > traffic. The Cisco 1720 does not translate any traffic where the
> > > destination can only be reached by using the default route. If there
is
>a
> > > specific route for the destination in the 1720 routing table, the 1720
> > > correctly translates and passes the traffic. If I originate traffic
>that
> > > must use the default route from the 1720, the 1720 routes it correctly.
> > > Now, I'm aware that the NAT will not work is there isn't a route to the
> > > destination in the routing table, or if there are access lists blocking
>a
> > > port. There are no access-lists in use anywhere in this scenario,
other
> > > than the one associated with NAT and the default route works correctly
> >when
> > > NAT isn't involved.
> > > Curiously, and this may or may not be important, I notice that I get a
> > > response to a 1.1.1.x address, but the router is only translating one
> > > way. It appears, oddly enough, that the router is generating the ICMP
> >echo
> > > response for the (virtual) 1.1.1.x address.
> > > Is there an issue with double-NAT of which I'm not aware?
> > >
> > > Thoughts?
> > >
> > > Craig
> > > FAQ, list archives, and subscription info:
> >http://www.groupstudy.com/list/cisco.html
> > > Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
> >FAQ, list archives, and subscription info:
> >http://www.groupstudy.com/list/cisco.html
> >Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>FAQ, list archives, and subscription info:
>http://www.groupstudy.com/list/cisco.html
>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=5931&t=5752
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]