Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: Re: [OMPI devel] openmpi-1.7.2 fails to use the RDMACM CPC
From: Steve Wise (swise_at_[hidden])
Date: 2013-08-19 16:40:17


> > I could if I had a patch/fix. :) I don't (yet) understand why HAVE_IBV_LINK_LAYER_ETHERNET was
> added.
> > Can the developer who made these changes explain the intent? I think it might have to do with
RoCE
> > support.
> >
>
> Seems like there should be some change to configure for adding this #define...

This commit added the new #define:

r27212 | jsquyres | 2012-08-31 18:42:37 -0700 (Fri, 31 Aug 2012) | 22 lines

Per some discussions between LANL, Cisco, ORNAL, and Mellanox, move
some new common OpenFabrics functionality to ompi/mca/common/verbs.
Also move everything that was in ompi/mca/common/ofautils under
ompi/mca/common/verbs.

 * Move ofautils -> verbs
 * Add new functionality in ompi/mca/common/verbs (see doxygen
 * comments in ompi/mca/common/verbs/common_verbs.h for details):
   * ompi_common_verbs_find_ibv_ports()
   * ompi_common_verbs_port_bw()
   * ompi_common_verbs_mtu()
   * '''If you're writing verbs-based code, you should be using this
     common functionality'''
 * Adapt openib BTL to use some trivial common functionality in
   common/verbs
 * Don't use "#ifdef OMPI_HAVE_RDMAOE",use
   "#if defined(HAVE_IBV_LINK_LAYER_ETHERNET)"
 * Update the following to include/link against common/verbs
   * bcol/iboffload
   * sbgp/ibnet
   * btl/openib

>
> > >
> > > On Aug 19, 2013, at 4:17 PM, Steve Wise <swise_at_[hidden]> wrote:
> > >
> > > >> -----Original Message-----
> > > >> From: devel [mailto:devel-bounces_at_[hidden]] On Behalf Of Steve Wise
> > > >> Sent: Monday, August 19, 2013 2:42 PM
> > > >> To: 'Open MPI Developers'; 'Jeff Squyres (jsquyres)'
> > > >> Cc: 'Indranil Choudhury'
> > > >> Subject: Re: [OMPI devel] openmpi-1.7.2 fails to use the RDMACM CPC
> > > >>
> > > >> I confirmed that this is a regression from 1.7.1...
> > > >>
> > > >> I'll see if I can figure out what's going on...
> > > >>
> > > >
> > > >
> > > > Looks like this is not defined anywhere: HAVE_IBV_LINK_LAYER_ETHERNET, which causes
> > > > btl_openib_connect_oob.c:oob_component_query() to falsely claim oob support for iwarp
devices.
> > > >
> > > > In 1.7.1 we see this in oob_component_query():
> > > >
> > > > #if defined(HAVE_STRUCT_IBV_DEVICE_TRANSPORT_TYPE)
> > > > if (BTL_OPENIB_CONNECT_BASE_CHECK_IF_NOT_IB(btl)) {
> > > > opal_output_verbose(5, mca_btl_base_output,
> > > > "openib BTL: oob CPC only supported on InfiniBand; skipped on
> %s:%d",
> > > > ibv_get_device_name(btl->device->ib_dev),
> > > > btl->port_num);
> > > > return OMPI_ERR_NOT_SUPPORTED;
> > > > }
> > > > #endif
> > > >
> > > > In 1.7.2, it adds the HAVE_IBV_LINK_LAYER_ETHERNET define:
> > > >
> > > > #if defined(HAVE_STRUCT_IBV_DEVICE_TRANSPORT_TYPE) &&
> > > defined(HAVE_IBV_LINK_LAYER_ETHERNET)
> > > > if (BTL_OPENIB_CONNECT_BASE_CHECK_IF_NOT_IB(btl)) {
> > > > opal_output_verbose(5, mca_btl_base_output,
> > > > "openib BTL: oob CPC only supported on InfiniBand; skipped on
> %s:%d",
> > > > ibv_get_device_name(btl->device->ib_dev),
> > > > btl->port_num);
> > > > return OMPI_ERR_NOT_SUPPORTED;
> > > > }
> > > > #endif
> > > >
> > > >
> > >
> > >
> > > --
> > > Jeff Squyres
> > > jsquyres_at_[hidden]
> > > For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
>
> _______________________________________________
> devel mailing list
> devel_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/devel