[strongSwan] Windows 7 seems to drop connection when rekeying main mode SA's

Wed Jan 11 09:51:09 CET 2012

I have now investigated things further..

CHILD_SAs is rekeyed successfully.
IKE_SAs however is not. This is what happens when the Strongswan host
(Debian, strongswan v4.5.2) is initiating a rekey of the IKE SAs.

Jan 11 09:33:29 firewall charon: 09[IKE] <rw-win-7|3> queueing IKE_REKEY task
Jan 11 09:33:29 firewall charon: 09[IKE] <rw-win-7|3> activating new tasks
Jan 11 09:33:29 firewall charon: 09[IKE] <rw-win-7|3>   activating
IKE_REKEY task
Jan 11 09:33:29 firewall charon: 09[IKE] <rw-win-7|3> IKE_SA
rw-win-7[3] state change: ESTABLISHED => REKEYING
Jan 11 09:33:29 firewall charon: 09[IKE] <rw-win-7|3> initiating
IKE_SA rw-win-7[4] to 82.147.51.146
Jan 11 09:33:29 firewall charon: 09[IKE] <rw-win-7|3> IKE_SA
rw-win-7[4] state change: CREATED => CONNECTING
Jan 11 09:33:29 firewall charon: 14[IKE] <rw-win-7|3> received DELETE
for IKE_SA rw-win-7[3]
Jan 11 09:33:29 firewall charon: 14[IKE] <rw-win-7|3> deleting IKE_SA
rw-win-7[3] between 77.106.149.54[C=NO, ST=Oppland, O=marsboer.net, O
                   U=VPN server,
CN=vpn.marsboer.net]...82.147.51.146[C=NO, ST=Oppland, O=marsboer.net,
OU=Roadwarriors, CN=rw01.marsboer.net]
Jan 11 09:33:29 firewall charon: 14[IKE] <rw-win-7|3> IKE_SA
rw-win-7[3] state change: REKEYING => DELETING
Jan 11 09:33:29 firewall charon: 14[IKE] <rw-win-7|3> IKE_SA deleted
Jan 11 09:33:29 firewall charon: 14[IKE] <rw-win-7|4> IKE_SA
rw-win-7[4] state change: CONNECTING => DESTROYING
Jan 11 09:33:29 firewall charon: 14[IKE] <rw-win-7|3> IKE_SA
rw-win-7[3] state change: DELETING => DESTROYING

I also tried with and without reauth and it did not change the results.
Withouth the strongswan host rekeying it seems like the Strongswan IKE
SA just dies, and then the connection has to be reestablished.
As soon as I reconnect the link is up again (done automatically after
1 minute on my Windows 7 host)

This is the network configuration:
Win 7 client (DHCP) -> NAT (hotspot fw, pfsense) -> NAT (main
firewall, Cisco ASA) -> Internet -> VPN server (dyndns)

Neither the DHCP address on the client or VPN server changed during my tests.
This is the same path that one of my Strongswan servers also uses so
it should be IKEv2 IPsec capable.

This is my current configuration:

# ipsec.conf - strongSwan IPsec configuration file

# basic configuration
config setup
        charonstart=yes
        plutostart=no

# Add connections here.
conn %default
        keyexchange=ikev2
        auth=esp
        authby=pubkey
        mobike=yes
        left=%defaultroute
        leftauth=pubkey
        leftcert=vpn-serverCert.pem
        leftfirewall=no
        type=tunnel
        dpdaction=clear
        dpddelay=300s

conn rw-uranus
        leftsubnet=10.10.10.0/24,10.0.1.0/24,10.10.99.0/24,10.0.2.0/24
        right=%any
        rightsourceip=10.0.1.2
        rightid="C=NO, ST=Oppland, O=marsboer.net, OU=Backup server,
CN=uranus.marsboer.net"
        auto=add
        ike=aes256-aesxcbc-ecp521
        esp=aes256gcm16-ecp521
        reauth=no

conn rw-win-7
        leftsubnet=0.0.0.0/0
        right=%any
        rightsourceip=10.0.1.0/24
        rightid="C=NO, ST=Oppland, O=marsboer.net, OU=Roadwarriors,
CN=rw01.marsboer.net"
        auto=add
        esp=aes256-sha1
        ikelifetime=90m
        reauth=no

conn rw-europa
        leftsubnet=10.10.10.0/24,10.0.1.0/24,10.10.99.0/24,10.0.2.0/24
        right=%any
        rightsourceip=10.0.1.4
        rightid="C=NO, ST=Oppland, O=marsboer.net, OU=VPN fileserver,
CN=europa.marsboer.net"
        auto=add
        ike=aes256-aesxcbc-ecp521
        esp=aes256gcm16-ecp521
        reauth=no

Regards,

Hans-Kristian Bakke

On Tue, Jan 10, 2012 at 17:16, Hans-Kristian Bakke <hkbakke at gmail.com> wrote:
> And one more thing:
> The "Rekey"-document explains that the CHILD_SA rekey timer should be
> more than 58m 46s if behind NAT and rekey active. This is the case in
> my situation.
> However, the configuration examples does not do anything to mitigate this.
>
> When the connection is newly started I can see that Strongswan wants
> to rekey CHILD_SA in about 48m (or around that).
>
> Does this mean that I have to set the "lifetime" parameter for the
> connection for i.e 90m (the default is 1h)
>
>
> Mvh
>
> Hans-Kristian Bakke
> Mob: 91 76 17 38
>
>
>
> On Tue, Jan 10, 2012 at 17:11, Hans-Kristian Bakke <hkbakke at gmail.com> wrote:
>> Thank you for your response
>>
>> I have read that document and have more or less based my config on it.
>> I have a couple of questions though:
>>
>> rekey is not mentioned in the X.509 example but is disabled in the
>> EAP-MSCHAP example. I have now reactivated rekey in my configuration
>> to test.
>>
>> I have set reauth to no because it made my strongswan to strongswan
>> tunnel drop the connection for a short moment. It is not mentioned in
>> the Windows 7 configuration.
>> Will having it enabled (like in the config examples) cause drop outs
>> during IKE SA renegotiations like I get using only strongswan?
>>
>> I now have rekey = on and reauth = on (default) to be as identical to
>> the example configuration as possible.
>> I will try using ! if it doesn't work, but in my case it will cause
>> issues because it will override the ike/esp parameters in my other
>> connections (older mailing list post, something to do with me having
>> %any as right in all my connections if I remember correctly)
>>
>> When it happens again I will look at the logs. Do you want a
>> particular log level or will the Debian default charon syslog do?
>>
>> Regards,
>>
>> Hans-Kristian Bakke
>>
>>
>>
>> On Tue, Jan 10, 2012 at 15:32, Martin Willi <martin at strongswan.org> wrote:
>>> Hi,
>>>
>>>> After disabling rekeying for Windows 7 connection I got rid of most of
>>>> the reconnects caused by rekeying the SAs, but I still have one
>>>> annoying connection interruption left.
>>>
>>> When following the rules from [1], rekeying initiated by strongSwan
>>> works fine here.
>>>
>>>> But for some reason IP Security Monitor on Windows 7 reports 10800s as
>>>> main mode SA lifetime. Even if I change ikelifetime on the Strongswan
>>>> server to i.e 8 or 12h it is still 3h.
>>>
>>> I don't know if you can trust the IP Security Monitor, as it is mainly
>>> for IKEv1. Not sure if these 10800s are correct. Further, lifetimes are
>>> never negotiated in IKEv2, you can't change the behavior of Windows by
>>> defining an ikelifetime on strongSwan. It only changes the behavior of
>>> rekeying initiated locally.
>>>
>>>> Now, the problem isn't really the 3h interval, it's that all the
>>>> connections drop for a while until reconnect.
>>>
>>> Would be helpful to know exactly _what_ is happening every three hours.
>>> Does Windows trigger a rekey? Does it drop the CHILD_SA, close the
>>> IKE_SA? A strongSwan log output would be helpful.
>>>
>>>>         ike=aes256-sha1-modp1024
>>>>         esp=aes256-sha1
>>>
>>> I'd try to limit the proposal list to exactly these by appending a '!'.
>>> I'm not aware of any problems with our lengthy default proposal set, but
>>> just in case.
>>>
>>> Regards
>>> Martin
>>>
>>> [1]http://wiki.strongswan.org/projects/strongswan/wiki/Windows7#Rekeying-behavior
>>>