[openib-general] OFED SRP initiator always sends CM REJ in response to CM REP

[Rimmer, Todd]

From: Chris Youb
Sent: Tuesday, October 31, 2006 12:13 PM
To: openib-general at openib.org
Subject: [openib-general] OFED SRP initiator always sends CM REJ in
response to CM REP

Abstract: 
 We are developing SRP target code and testing it with the OFED 1.1 SRP
initiator.  The OFED SRP initiator sends us a CM REQ (IB 12.6.5) and we
respond with CM REP (12.6.8).  However, instead of the expected CM RTU
(12.6.9) we ALWAYS receive a CM REJ (12.6.7) with status 0x1C == Reason
28 (12.6.7.2). 

Software Setup: 
- SUSE 10.0 
- OFED 1.1 
- Mellanox card with 3.5.00 firmware 

Details: 
 Initially we suspected our response values in the CM REP packet.  There
is nothing obvious to us, and anything we weren't sure about we tried a
number of combinations.  This applies to the SRP private data as well. 
 We also took a look at ./openib-1.1/drivers/infiniband/core/ cm.c and
cma.c.  The function cma.c:cma_rep_recv looked like a possibility but
there's little debug output.  Aside from putting in printk's and
recompiling and installing is there an easier way to debug? 

 
Chris,
 
I reviewed the packets and found at least 1 problem.
 
The REQ has responder resources=0x4, yet the REP has initiator depth=7.
The REP must provide an initiator depth <= the responder resources in
the REQ.
 
Some other non-fatal issues:
Target Ack Delay is a bit high (0x1f -> 2.4 hours).  This will basically
cause initiator to compute QP Ack timeouts of 2.4 hours.  This value
should represent the time internal to the CA from receipt of a message
to sending of the ACK.  See IBTA 12.7.33 for more info.
 
Todd Rimmer
 
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20061031/3a0110af/attachment.html>