[Openswan Users] traffic only being encrypted one way

Bob Benstro bbenstro at gmail.com
Tue Mar 20 22:15:17 EDT 2007


Paul / Harald,

Finally managed to get a hub hooked up today.  The packets I'm now seeing
are PPPOE encapsulated, so, I know I'm looking at things via tcpdump after
it's left for the modem.. and openswan should have done its work.

I'm seeing packets routed for 192.168.15.90 over the net, as I described
before. :/  Something is definitely wonky here.  Just to be sure, I removed
all iptables rules, removed all routes, and started with a fresh routing
table.  I brought up one PPPOE interface / modem, restarted IPSEC, and went
from there.

Packets left via the correct route, but were still PPPOE encapasuated but
not IPSEC encrypted.

On this box I'm using the routing patches described on the
http://lartc.orgadvanced routing howto (near the bottom of this page):

http://lartc.org/howto/lartc.rpdb.multiple-links.html#AEN298

The link points to here:

http://www.ssi.bg/~ja/#routes

I'm using kernel  2.6.17.7.

I have 4 cable modems and 2 DSL modems hooked up in this configuration..
however, so far Openswan has worked flawlessly using l2tpd.  It is only when
I have gone into a linux to linux setup that problems exist.  For example, I
have one l2tp connected host up right now, and I can ping it without issue.

Suggesting that routing might be mucking with things, or NAT rules makes
sense.. as all of my  l2tpd routed openswan connections use non-private IP
space in the routing table, and all of my openswan to openswan connections
use private subnet routes/addresses.  Still, I flushed all routes via ip
route flush, and iptables via -F, and confirmed they had vanished.  This
should not be the cause, then.

Roadwarrior side:

config setup
        nat_traversal=yes
        nhelpers=0

include /etc/ipsec.d/examples/no_oe.conf

conn rw
    dpdaction=restart
    dpdtimeout=120
    authby=rsasig
    pfs=no
    keyingtries=3
    left=%defaultroute
    leftcert=/etc/ipsec.d/rw.pem
    leftrsasigkey=%cert
    right=1.1.1.1
    rightcert=base.pem
    rightsubnet=192.168.0.0/24
    auto=start

The 'base' has a more complex configuration, naturally:

config setup
        nat_traversal=yes
        virtual_private=%v4:
10.0.0.0/8,%v4:172.16.0.0/12,%v4:192.168.0.0/16,%v4:!192.168.2.0/24

conn %default
        compress=yes
        authby=rsasig
        pfs=no
        leftcert=base.pem
        right=%any
        rightca=%same
        rightrsasigkey=%cert
        rightsubnet=vhost:%no,%priv
        auto=add
        keyingtries=3

#Disable Opportunistic Encryption
include /etc/ipsec.d/examples/no_oe.conf

conn internal
        left=192.168.0.1
        leftprotoport=17/1701
        rightprotoport=17/1701

conn cable1
        left=3.3.3.3
        leftnexthop=3.3.3.4
        leftprotoport=17/1701
        rightprotoport=17/1701

conn pppoe1
        left=1.1.1.1
        leftnexthop=1.1.1.2
        leftprotoport=17/1701
        rightprotoport=17/1701

conn pppoe2
        left=2.2.2.2
        leftnexthop=2.2.2.3
        leftprotoport=17/1701
        rightprotoport=17/1701

conn internal10
        left=10.10.10.1
        leftprotoport=17/1701
        rightprotoport=17/1701

conn rw
        auto=add
        left=1.1.1.1
        leftnexthop=1.1.1.2
        leftsubnet=192.168.0.0/24
        rightnexthop=%defaultroute
        rightcert=rw.pem


I'm not sure where else to go from here.  Keep in mind, as review, the
connection does come up and works perfectly from the road warrior side.  I
can connect to sendmail, imap, you name it .. but I am unable to get any
packets initiated from the base to the road warrior...

Thanks (I hope!! ;)



On 3/16/07, Bob Benstro <bbenstro at gmail.com> wrote:
>
>
>
> On 3/16/07, Paul Wouters <paul at xelerance.com> wrote:
> >
> > On Fri, 16 Mar 2007, Bob Benstro wrote:
> >
> > > > Most often this is due to the vpn server not being the default
> > gateway, and
> > > > the local subnet sending the traffic for the vpn to the default
> > gateway,
> > > > instead of the vpn server.
> > > >
> > > >
> > > I'm not sure what you mean.  It seems weird that you've removed from
> > my
> > > quoted material above, the text that provides information showing this
> > isn't
> > > the case.
> >
> > I did not see that information in your email.
> >
> > > Anyhow, as I mentioned, the traffic is indeed leaving the correctly
> > routed
> > > interface as it should be.   The only problem is that the traffic
> > leaving
> > > that interface is not encrypted.  It is, however, leaving the
> > interface it
> > > should be leaving, in order to reach the remote box.   My local subnet
> > and
> > > its default route is not in question, as I am performing all tests on
> > the
> > > VPN box itself, so no need to worry there.
> >
> > Okay. Are you sure the traffic leaves unencrypted? If you use KLIPS,
> > that is
> > indeed easy to see, just compare outgoing physical interface with ipsecX
> > interface. With NETKEY, you don't get to see the encrypted packets
> > before they
> > leave your box, they are encrypted AFTER tcpdump can see them, so this
> > cannot
> > be proven using the sending box.
>
>
>
> Hmm.  Well, I'm using NETKEY, I haven't patched my kernel or anything of
> the sort.  There is no ipsecX device for me.
>
> However, on the remote box, I can definitely see encrypted packets,
> although I suppose these could merely be packets returning when I am pinging
> or otherwise.  This lead me to believe that I should be able to see
> encrypted packets, but fair enough.
>
>
> Since if they were cleartext, they would go
> > to some unknown private space and get dropped, you cannot see it on the
> > receiving
> > end either. But you might see encrypted packets arriving on the
> > receiving end,
> > which are never successfully decrypted for some reason (NAT, ipsec
> > passthrough
> > corruption, etc).
>
>
>
> No, unfortunately there is absolutely no incoming traffic on the remote
> box that I can see, when pinging/etc from the local to remote box, encrypted
> or otherwise. :/
>
>
> Then there is also the possibility you are in fact sending out encrypted
> > ESP
> > packets (which you can't see when using NETKEY), but some filter
> > somewhere filters
> > the ESP packets and they never arrive at the destination. Again, you
> > would
> > not be able to easilly distinguish this from the case they are never
> > encrypted,
> > send to a bogus router and dropped.
>
>
>
> This could indeed be the case... but I suppose I would need to hookup a
> hub and another box to watch for said case?  Can you think of an easier way?
>
> Right now, if I tracedump to the remote box outside of the openswan setup
> (direct external IP to external IP), I get a successful traceroute of about
> 9 hops, ending at the remote box.  If I tracedump using the extruded IP from
> the remote box, it drops on the floor after 4 hops, which could support your
> theory of a router dropping them along the way.  Blech. :/
>
>
>
> This is why I asked for more information. Knowing whether you use KLIPS or
> > NETKEY
> > on the sending end would help reduce the possible scenarios.
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.openswan.org/pipermail/users/attachments/20070320/fe714429/attachment-0001.html 


More information about the Users mailing list