Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: [OMPI devel] PATCH: Wrong event_type value passed in to show_help when getting xrc async events
From: nadia.derbey (Nadia.Derbey_at_[hidden])
Date: 2010-07-13 01:58:56


Hi,

When the asynchronous events device handler(btl_openib_async_deviceh())
gets an async event and XRC is enabled, the XRC bit is cleared to
process the event_type value, but orte_show_help is called with the
event_type original value (i.e. XRC bit still present). This leads to
the following kind of message:

  ----------------------------------------------------------
  The OpenFabrics stack has reported a network error event. Open MPI
  will try to continue. but your job may end up failing.

   Local host: XXXX
   MPI process PID: 31818
   Error number: -2147483645 (UNKNOWN)
  This error may indicate connectivity problems within the fabric;
  please contact your system administrator
  -----------------------------------------------------------

While the expected error number is
  Error number: 3 (IBV_EVENT_QP_ACCESS_ERR)

I propose the attached small patch to fix this issue.

Regards,
Nadia

-- 
nadia.derbey <Nadia.Derbey_at_[hidden]>