[openib-general] troubles with IPoIB

[Josh England]

Hi all,

I've got an 85-node x86_64 PCIe cluster I'd like to run (and test)
openIB on.  I've built a kernel using the latest patches from SVN,
loaded all the modules, and I see ACTIVE on the ports, but IPoIB does
not seem to want to work.

After I 'echo 2 > module/ib_ipoib/debug_level', I get the following:

Nov 22 16:47:03 n0 kernel: ib0: called: id 14, op 0, status: 0
Nov 22 16:47:03 n0 kernel: ib0: send complete, wrid 14
Nov 22 16:47:03 n0 kernel: ib0: called: id -2147483634, op 128, status:
0
Nov 22 16:47:03 n0 kernel: ib0: received 100 bytes, SLID 0x000d
Nov 22 16:47:03 n0 kernel: ib0: dropping loopback packet
Nov 22 16:47:04 n0 kernel: ib0: sending packet, length=60
address=000001007fd17340 qpn=0xffffff

Is the 'called: id [big-fat-negative-number]' supposed to be there?  Is
there a utility (such as vping) to test even basic IB connectivity?  How
can I get IPoIB to work?

-----------------------------------------------
Josh England
Sandia National Laboratory, Livermore, CA
Visualization and Scientific Computing
email: jjengla at sandia.gov
phone: (925) 294-2076

-JE