[openib-general] librdmacm and udapl: Which git branch to use in ofed_1_2 build

[Michael S. Tsirkin]

> Quoting Sean Hefty <mshefty at ichips.intel.com>:
> Subject: Re: librdmacm and udapl: Which git branch to use in ofed_1_2 build
> 
> Michael S. Tsirkin wrote:
> >>Where will you branch release 1.0 from? Since we agreed OFED will be based
> >>on release 1.0 of librdmacm, this is what OFED shall take.
> > 
> > Sean, could you answer this please?
> 
> This depends on the status of the multicast changes by feature freeze, and which 
> kernel OFED wants to ship.  Since multicast support has been requested, I would 
> like to get this in.  However, the kernel multicast changes are not in 2.6.20. 
>    What's OFED's stance on kernel patches on this?  This is different than the 
> local_sa, which is mostly self contained, since the multicast patches affect ipoib.

Yes, but everyone agreed to have this patch in (that is - several
people want it and no one objects).  As far as Mellanox testing goes, crashes we
saw with it were resolved.

Again, I am just trying to avoid duplicating work, so I am waiting for
you to tell me when to take the code and put in OFED.

> For now, current branch (rdma_ucm-abi3) that OFED is pulling is fine.

OK. Note however we won't be able to switch after feature freeze,
so whatever it will be in OFED needs to be decided by then.

> On a side note, I'm not sure that the release should be called 1.0.  When OFED 
> 1.1 shipped the librdmacm, there was not a release number assigned, and my 
> understanding is that the distributors assigned one.  Can we clarify how we want 
> to deal with this?

I think we can just ignore this.

OFED 1.1 shipped a pre-release and named it (I think) 1.0, but I am not sure
this means we have to worry about collisions too much - this is always a problem
with distributing pre-releases.

And note how once kernel part is updated an old librdmacm lib will break anyway
even if you do not override it.

-- 
MST