[openib-general] Debugging Infiniband?

[Walls, Jeffrey Joel]

Hi Dotan,

Thank you for your suggestions!  I was finally able to figure out my
problem.  I was
Using the WinIB package from Mellanox, and there were some fields in the
data
Structures that were in big-endian format, even though it was running on
Win32.  Once
I got these figured out and swapped correctly, things worked again.

Thanks again!

Jeff

-----Original Message-----
From: Dotan Barak [mailto:dotanb at mellanox.co.il] 
Sent: Wednesday, January 18, 2006 8:53 AM
To: Walls, Jeffrey Joel; openib-general
Subject: RE: [openib-general] Debugging Infiniband?

Hi jeff.
 
 
there are some issues you need to check:
there are WR that were posted to the remote QP RQ before posting the WR
to the SQ in local side both of the QPs are alive and in valid states
(at least RTR for responder and RTS for requestor) the QPs parameters
are synch (for example: the psn) the route that you are using is valid
(port, remote QP number, remote lid) if you are using UD/UC QPs maybe
the packet were dropped ..
 
if you have an IB analyzer you should check that the packet was sent to
the expected QP number you can check the port counters to see how many
data was sent / received to each IB port
 
 
I Hope i gave you some useful information

[Dotan Barak] 

	 -----Original Message-----
	From: openib-general-bounces at openib.org
[mailto:openib-general-bounces at openib.org]On Behalf Of Walls, Jeffrey
Joel
	Sent: Wednesday, January 18, 2006 5:39 PM
	To: openib-general
	Subject: [openib-general] Debugging Infiniband?
	
	
	Hi,
	 
	I first must admit that I'm new to Infiniband and Infiniband
programming.
	I have just begun writing my first commercial application using
IB late last
	year.  I'm very familiar with socket programming (TCP,
Multicast, etc), though.
	 
	I'm wondering what techniques expierenced IB programmers use to
debug
	IB applications.
	 
	My situation is that I'm running an data producer on Windows XP
and a set
	of data consumers on Linux.  So for Windows, I'm using WinIB
(gen1) and
	for linux I'm using OpenIB (gen2).
	 
	I have both sides implemented according to some of the example
code I've
	seen and also according to the documents I've been able to find.
The connections
	all seem to be set up properly and my producer successfully
posts all of its sends
	(at least according to my CQE's returned).
	 
	The problem is that my receiver never sees any of the IB
packets.  I post the receive
	and then wait forever polling the CQ.
	 
	I've run out of ideas on what to even look at and am now looking
for suggestions on
	how to best figure out this problem.
	 
	If you have any ideas or need more clarification, I'd love to
hear from you.
	 
	Also, if this isn't the proper forum for such discussions, if
you could please guide me
	in the right direction, I would greatly appreciate that as well.
	 
	Best Regards,
	Jeff