[strongSwan] Multiple CHILD_SA problem

[John Brown]

Hi all,

 I am facing some problems with strongswan 4.5.2 or 5.2.1 (currenty tested)
on debian wheezy (armel). One of these problems is having multiple CHILD_SA
created under Security Association

 For example, fragment of the output from "ipsec statusall" taken from
remote device looks like this:

        Security Associations (1 up, 0 connecting):
   vpn1[17]: ESTABLISHED 33 minutes ago, 2001:db8:de5c::134[C=AA,
O=Test-co, CN=dev1]...2001:db8:cab1::201[C=AA, O=Test-co, CN=dev2]
   vpn1[17]: IKEv2 SPIs: 6e7520807913c57c_i* 404ad97e8447df4a_r, rekeying
in 60 minutes, public key reauthentication in 2 hours
   vpn1[17]: IKE proposal: AES_GCM_16_256/PRF_HMAC_SHA2_512/MODP_2048
   vpn1{10}:  INSTALLED, TUNNEL, ESP SPIs: c947dd26_i c85fc3c0_o
   vpn1{10}:  AES_GCM_12_256, 2496 bytes_i (24 pkts, 0s ago), 3120 bytes_o
(30 pkts, 0s ago), rekeying in 9 minutes
   vpn1{10}:   2001:db8:de5c::134/128 === 2001:db8:cab2::/64
   vpn1{10}:  INSTALLED, TUNNEL, ESP SPIs: c04a385a_i cdcc2a4b_o
   vpn1{10}:  AES_GCM_12_256, 23296 bytes_i (224 pkts, 0s ago), 23920
bytes_o (230 pkts, 0s ago), rekeying in 7 minutes
   vpn1{10}:   2001:db8:de5c::134/128 === 2001:db8:cab2::/64
   vpn1{10}:  INSTALLED, TUNNEL, ESP SPIs: c09d50d8_i c72fefde_o
   vpn1{10}:  AES_GCM_12_256, 16016 bytes_i (154 pkts, 0s ago), 16640
bytes_o (160 pkts, 0s ago), rekeying in 10 minutes
   vpn1{10}:   2001:db8:de5c::134/128 === 2001:db8:cab2::/64
   vpn1{10}:  INSTALLED, TUNNEL, ESP SPIs: ce87de01_i c0bd6c13_o
   vpn1{10}:  AES_GCM_12_256, 23088 bytes_i (222 pkts, 0s ago), 23816
bytes_o (229 pkts, 0s ago), rekeying in 8 minutes
   vpn1{10}:   2001:db8:de5c::134/128 === 2001:db8:cab2::/64
   vpn1{10}:  INSTALLED, TUNNEL, ESP SPIs: c83839f6_i c5bee9c4_o
   vpn1{10}:  AES_GCM_12_256, 2392 bytes_i (23 pkts, 0s ago), 3120 bytes_o
(30 pkts, 0s ago), rekeying in 10 minutes
   vpn1{10}:   2001:db8:de5c::134/128 === 2001:db8:cab2::/64
   vpn1{10}:  INSTALLED, TUNNEL, ESP SPIs: c4f3ff51_i ce1cb4f4_o
   vpn1{10}:  AES_GCM_12_256, 40872 bytes_i (393 pkts, 0s ago), 41600
bytes_o (400 pkts, 0s ago), rekeying in 9 minutes
   vpn1{10}:   2001:db8:de5c::134/128 === 2001:db8:cab2::/64
   vpn1{10}:  INSTALLED, TUNNEL, ESP SPIs: cce7630b_i c125f156_o
   vpn1{10}:  AES_GCM_12_256, 10712 bytes_i (103 pkts, 0s ago), 11440
bytes_o (110 pkts, 0s ago), rekeying in 11 minutes
   vpn1{10}:   2001:db8:de5c::134/128 === 2001:db8:cab2::/64
   vpn1{10}:  INSTALLED, TUNNEL, ESP SPIs: c106eb4e_i cdeae2d2_o
   vpn1{10}:  AES_GCM_12_256, 34528 bytes_i (332 pkts, 0s ago), 35152
bytes_o (338 pkts, 0s ago), rekeying in 9 minutes
   vpn1{10}:   2001:db8:de5c::134/128 === 2001:db8:cab2::/64
   vpn1{10}:  INSTALLED, TUNNEL, ESP SPIs: cc5b9dd0_i c3954198_o
   vpn1{10}:  AES_GCM_12_256, 662792 bytes_i (6373 pkts, 0s ago), 662792
bytes_o (6373 pkts, 0s ago), rekeying in 10 minutes
   vpn1{10}:   2001:db8:de5c::134/128 === 2001:db8:cab2::/64


   This is after 22 hours of working tunnel in which for test purposes goes
icmp traffic (~10 requests/sec). As the time goes by, the number of
CHILD_SA created grows up. There is of course only one CHILD_SA just after
the tunnel is established.
   I could not find a solution but I've found an option "inactivity" which
"defines the timeout interval, after which a CHILD_SA is closed if it did
not send or receive any traffic". I've got other device with sswan 4.5.2
(wheezy, armel) with inactivity set to 3m and there I do not observe many
CHILD_SA now, but here they are. As I execute "ipsec statusall" every few
seconds I see that only one CHILD_SA is used for traffic at a time and
others have unchanged number of bytes transferred but the time elapsed
since particular CHILD_SA was used is very often 0s after execute "ipsec
statusall" (as you can see above) and is increasing after subsequent calls
but never reaches the 5m, it falls back to 0 after no more than ~200seconds.

   The remote device was an ikev2 initiator.

   Does anyone know what might be causing the problem?

   Normally I am using longer key life times but they were shortened during
the tests which probably increase the number of CHILD_SA in a particular
period of time.

   Configurations:

   remote device config:

conn vpn1
        left=2001:db8:de5c::134
        right=2001:db8:cab1::201
        rightsubnet=2001:db8:cab2::201/64
        ike=aes256gcm12-aes256-sha256-modp2048!
        esp=aes256gcm12-aes256-sha256-modp2048!
        keyexchange=ikev2
        ikelifetime=100m
        lifetime=30m
        inactivity=5m
        reauth=no
        rekeymargin=300s
        rekeyfuzz=100%
        mobike=no
        auto=route
        reqid=10
        leftauth=pubkey
        rightauth=pubkey
        leftcert=d1.crt
        rightid="C=AA, O=Test-co, CN=dev2"
        leftid="C=AA, O=Test-co, CN=dev1"



gate config:

   conn toClient1-ipv6
        left=2001:db8:cab1::201
        leftsubnet=2001:db8:cab2::/64
        right=2001:db8:de5c::134
        ike=aes256gcm16-sha512-modp2048!
        esp=aes256gcm12-sha256-modp2048!
        keyexchange=ikev2
        ikelifetime=180m
        lifetime=60m
        rekeymargin=300s
        rekeyfuzz=100%
        mobike=no
        auto=route
        leftauth=pubkey
        rightauth=pubkey
        rightid="%any"
        leftcert=gate.crt
        reqid=10
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.strongswan.org/pipermail/users/attachments/20160226/2d029a9c/attachment.html>