This is useful for kernel implementations where the ordering of SAs
is unpredictable and the new SA might otherwise not be used until the
DELETE response has been received, which is not ideal as the responder
might not keep the old SA around that long. On Linux, it makes no
difference as we switch to the new outbound SA immediately because the
updated outbound policy references its SPI.
The peer might not have seen the CREATE_CHILD_SA response yet, receiving a
DELETE for the SA could then trigger it to abort the rekeying, causing
the deletion of the newly established SA (it can't know whether the
DELETE was sent due to an expire or because the user manually deleted
it). We just treat this SA as if we received a DELETE for it. This is
not an ideal situation anyway, as it causes some traffic to get dropped,
so it should usually be avoided by setting appropriate soft and hard limits.
References #2815.
This splits the SA installation also on the initiator, so we can avoid
installing the outbound SA if we lost a rekey collision, which might
have caused traffic loss depending on the timing of the DELETEs that are
sent in both directions.
After deleting a rekeyed CHILD_SA we uninstall the outbound SA but don't
destroy the CHILD_SA (and the inbound SA) immediately. We delay it
a few seconds or until the SA expires to allow delayed packets to get
processed. The CHILD_SA remains in state CHILD_DELETING until it finally
gets destroyed.
The responder has all the information needed to install both SAs before
the initiator does. So if the responder immediately installs the outbound
SA it might send packets using the new SA which the initiator is not yet
able to process. This can be avoided by delaying the installation of the
outbound SA until the replaced SA is deleted.
Tobias Brunner