[strongSwan] how to increase timeout for "deleting half open IKE_SA with <someIP> after timeout" ?

[Thomas Egerer]

Hi Harri,


On 5/14/21 11:06 AM, Harald Dunkel wrote:
> Hi folks,
>
> I have a few road warriors (3 out of ~140) having severe problems to
> connect via IKEv2. Within the last 4 weeks they had >1000 problems
> during IKE SA init each, e.g.:
>
> May 12 09:55:28 18[NET1] <92244> received packet: from 192.168.1.177[61416] to 10.0.0.17[500] (432 bytes)
> May 12 09:55:28 18[ENC1] <92244> parsed IKE_SA_INIT request 0 [ SA KE No N(REDIR_SUP) N(NATD_S_IP) N(NATD_D_IP) N(FRAG_SUP) ]
> May 12 09:55:28 18[IKE0] <92244> 192.168.1.177 is initiating an IKE_SA
> May 12 09:55:28 18[IKE2] <92244> IKE_SA (unnamed)[92244] state change: CREATED => CONNECTING
> May 12 09:55:28 18[CFG1] <92244> selected proposal: IKE:AES_CBC_256/HMAC_SHA2_256_128/PRF_HMAC_SHA2_256/MODP_2048
> May 12 09:55:28 18[IKE1] <92244> remote host is behind NAT
> May 12 09:55:28 18[IKE2] <92244> sending strongSwan vendor ID
> May 12 09:55:28 18[IKE1] <92244> sending cert request for "C=DE, O=example AG, CN=ws-CA"
> May 12 09:55:28 18[IKE1] <92244> sending cert request for "C=DE, O=example AG, OU=example Certificate Authority, CN=root-CA"
> May 12 09:55:28 18[IKE1] <92244> sending cert request for "C=DE, ST=NRW, O=example AG, OU=TI, CN=ipsec-ca"
> May 12 09:55:28 18[ENC1] <92244> generating IKE_SA_INIT response 0 [ SA KE No N(NATD_S_IP) N(NATD_D_IP) CERTREQ N(FRAG_SUP) N(CHDLESS_SUP) N(MULT_AUTH) V ]
> May 12 09:55:28 18[NET1] <92244> sending packet: from 10.0.0.17[500] to 192.168.1.177[61416] (541 bytes)
> May 12 09:55:58 31[JOB1] <92244> deleting half open IKE_SA with 192.168.1.177 after timeout
> May 12 09:55:58 31[IKE2] <92244> IKE_SA (unnamed)[92244] state change: CONNECTING => DESTROYING
>
> Obviously there is a 30sec timeout on the IPsec gateway. Is there
> a chance to increase this timeout (using stroke, ie. ipsec.conf)?
> https://wiki.strongswan.org/projects/strongswan/wiki/ConnSection
> mentions only the DPD timeout (150 sec per default) and the inac-
> tivity timeout (child sa only, as it seems).

What you're looking for is not a part of the swanctl.conf or ipsec.conf. It's part
of the strongswan.conf (/etc/strongswan.conf) and documented at [1]. Use
the keyword 'half_open_timeout' in the 'charon' section:

<snip>
charon {
	half_open_timeout = 42
}
<snap>

If it takes your RW peers more than 30 seconds to respond one would assume that
the response got lost along the way. Normally in such a case the initiator would
retransmit the request. That's not part of the responders job.

>
> Would it be wise to resend the IKE_SA_INIT response (lets say) 3
> times?
As mentioned above, you may check your RWs configuration to retransmit the request.
>
> Every helpful comment is highly appreciated
>
> Harri

Cheers,
Thomas

[1] https://wiki.strongswan.org/projects/strongswan/wiki/strongswanconf