I have come across some error codes which are not entirely documented
(or at least very well hidden)..
They are generated when I try to send via rdma using uverbs interface on
mthca drivers.
When polling for work completions, I get the following behavior, and
have not been able to figure out if they are related, and/or what they
actually mean.
( I'm polling using ibv_poll_cq() inside a loop)
I receive these two error values/codes for the ibv_wc.status variable:
IBV_WC_LOC_PROT_ERR
IBV_WC_WR_FLUSH_ERR (after the first one, all of the status fields are
this)
Is the LOC_PROT_ERR being generated due to some wrong access flags for
the queue pair upon initialization?
I can send/recv fine on the same queue pair when not using rdma, so I'm
somewhat confused about why I'd not get this error elsewhere.
what causes the two error codes?
thanks,
- Kyle
--
Kyle Schochenmaier
[EMAIL PROTECTED]
Research Assistant, Dr. Brett Bode
AmesLab - US Dept.Energy
Scalable Computing Laboratory
_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
