[Swan-dev] broken delete behaviour on ikev2 ?

[Paul Wouters]

Testing the ike header flags by sending an informational delete message,
I see an unrelated problem, the delete fails:

| **parse ISAKMP Message:
|    initiator cookie:
|   ec 52 33 3e  d5 21 42 5c
|    responder cookie:
|   73 41 2d e0  dd 4d 92 b3
|    next payload type: ISAKMP_NEXT_v2E
|    ISAKMP version: IKEv2 version 2.0 (rfc4306/rfc5996)
|    exchange type: ISAKMP_v2_INFORMATIONAL
|    flags: ISAKMP_FLAG_v2_IKE_INIT
|    message ID:  00 00 00 01
|    length: 76
|  processing version=2.0 packet with exchange type=ISAKMP_v2_INFORMATIONAL (37)
| I am receiving an IKE Request
| I am the IKE SA Original Responder
| ICOOKIE:  ec 52 33 3e  d5 21 42 5c
| RCOOKIE:  73 41 2d e0  dd 4d 92 b3
| state hash entry 1
| parent v2 peer and cookies match on #1
| v2 state object #1 found, in STATE_PARENT_R2
packet from 192.1.2.45:500: received too old retransmit: 1 < 2

This seems correct. For the informational exchange, we should have used
msgid 2 since we used 0 for INIT and 1 for AUTH.

In send_inforational() we use:

 	r_hdr.isa_msgid = htonl(pst->st_msgid_nextuse);

So it looks like sending a successfull IKE_AUTH reply packet did not bump
st_msgid_nextuse at the end of send_informational().

The msgid bump is handled in ikev2_update_counters() which is called by
success_v2_state_transition()

I'll add some logging to ikev2_update_counters() and investigate
tomorrow.

This must have shown up in our running testcases?  Antony, can you see
when this failure started?

Paul