[ofw] Re: Completion with bad status: IBV_WC_RETRY_EXC_ERROR

[Diego Guella]

Hi Tzachi,

The error is on the "daemon" side (sender) at the polling of the cq, I get a completion work, but the status is not success.
The "client" (receiver) don't sees nothing, in the recv case none of the recv wr are used. In the rdma case, it is normal that the recv wr isn't used.

I am using either send/recv and rdma operations but the results are the same.
I am doing only one operation, the code is the example in WinIB 1.3, under Mellanox\WinIB\SDK\examples\hello_world
with some slight modifications on Windows to get RDMA working, and heavy modifications on Linux to use the Verbs interface instead of ib_al.

I can send you the sources if you want.


The same program works OK on Windows using the same machine as client/daemon. (I don't have 2 windows machines with Infiniband here)



  ----- Original Message ----- 
  From: Tzachi Dar 
  To: Diego Guella ; Fab Tillier 
  Cc: ofw at lists.openfabrics.org 
  Sent: Tuesday, November 13, 2007 12:59 PM
  Subject: RE: Completion with bad status: IBV_WC_RETRY_EXC_ERROR


  Hi Diego,

  Can you give us more information on where the error is?

  Does your problem happen on the first packet that you send or only after some time?
  Are you using send/recv or RDMA operations?

  Thanks
  Tzachi



----------------------------------------------------------------------------
    From: Diego Guella [mailto:diego.guella at sircomtech.com] 
    Sent: Tuesday, November 13, 2007 11:29 AM
    To: Diego Guella; Tzachi Dar; Fab Tillier
    Cc: ofw at lists.openfabrics.org
    Subject: Completion with bad status: IBV_WC_RETRY_EXC_ERROR


    Hi

    I'm trying to port the source example in Mellanox\WinIB\SDK\examples\hello_world to Linux.

    The problem I have is this: after I post the WR, I get completion with error:
    status = 12 (IBV_WC_RETRY_EXC_ERROR) on Linux;
    status = IB_WCS_TIMEOUT_RETRY_ERR on Windows.


    What does this error means? I'm not sure about it.
    Perhaps am I sending to the wrong LID? to the wrong QPN??

    Any help is appreciated.


    Thanks,
    Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/ofw/attachments/20071113/f6f0428f/attachment.html>