(racoon2 50) Re: graceful shutdown

 In your previous mail you wrote:

   Francis Dupont wrote:
   > iked should delete/abort the IKE SA and flush IPsec SA when it is killed,
   > i.e., in iked/main.c terminate_iked() I'd like to see the comment:
   > 
   > 	/* flush IPsec SA ? */
   > 
   > replaces by the code as the answer to the question is YES!
   
   You mean with Informational exchange and DELETE payload?

=> yes, you have to send one per establish IKE SA even you should not
wait for an answer. The idea is to give the peer a chance to know you
are going down.

   I'll add it, but I was wondering what to do if a peer connection is down.

=> ignore any error of cause, i.e., just try!

   Perhaps,
   try DELETE for first SIGINT, and if SIGINT is received twice,
   immediately delete all SAs and exit?  Or, try DELETE with SIGTERM, and
   immediately exit with SIGINT?
   
=> no, I believe this is too graceful. There are already some mechanisms
to recover with a lost DELETE. BTW what is really annoying in tests
is to have to flush the IPsec SAs manualy, in particular on the initiator
(because you don't get ACQUIREs :-), on the responder you are in trouble
only with the preferred_oldsa.

Thanks

[email protected]