[openib-general] [PATCH] librdmacm: updated librdmacm to work with proposed 2.6.20 kernel CMA

[Or Gerlitz]

On 11/2/06, Sean Hefty <mshefty at ichips.intel.com> wrote:
> > Have you looked on that? from the compilation failure against
> > libibverbs-1.0 the gap seem pretty small. If indeed this is the case,
> > since libibverbs-1.1 is in development lets check with Roland if it
> > makes sense for him to support these small-gap-features in
> > libibverbs-1.0.X, i guess what matters here is ABI versions...
>
> I have not had time to look into this yet.

I see, this is of great importance, but in second priority i think to
the kernel push (we are now in post -rc4 time frame, which means
2.6.20 is going to get open within few weeks...).
However, if u/k ABI issues are involved here, we might get into
troubles... anyway, whenever you have some time...

> > I think we do want it. The rdma cm provide the means to offload ip
> > multicast to ib multicast though registration (join/leave etc) with the
> > ib_sa module. IP Multicast does use the send-only feature and hence IP
> > Multicast offloading apps need it as well. The rdma cm framework fits
> > very well for such apps and the ib_usa (which does not exist now, and i
> > am not sure needs to exist... it was a project of a summer student with
> > open-mpi that required that...) not.

> Are you wanting the rdma cm to join the same multicast groups that ipoib does?
> (This is simple to change, but it does not join the same groups today.)

Actually, yes, we strogly prefer that you will remove the rdma cm
signature byte from MGIDs generated by the cma.

> I will likely need to spin these patches again to incorporate the changes for
> path failover, so adding in join options wouldn't be difficult.  Are you just
> wanting to see them added the rdma_join_multicast directly?

Yes, i'd like to have the rdma cm consumer being able to specify if it
want to be joined as full member vs join as send only. This would also
have an impact on the rdma cm calling the qp attach to mgid verb for
the first case, only.

Or.