[openib-general] question related to rdma_bind_addr

[James Lentini]

On Thu, 23 Mar 2006, Sean Hefty wrote:

> >What does it mean to bind to a remote address? What functionality
> >would that enable? Spoofing?
> 
> I think that Or is just exploring the idea of synchronously binding 
> to a local *device* based on a remote address.  
>
> This would allow an application to bind, then allocate PDs, CQs, 
> QPs, etc. up front, rather than deferring resource allocation until 
> address resolution completes.  A ULP may be able to take advantage 
> of this, but I can't personally say that I know what benefit it 
> would provide.  (Maybe avoid the need to keep track of everything 
> that must be allocated once address resolution completes?)
>
> >When I think of bind(2), I only think of binding to local 
> >addresses.
> 
> Yes - this is what rdma_bind_addr(src_addr) does.  But I can 
> envision adding a new call, rdma_bind_device(dst_addr), provided 
> some use for it can be found.

I would find calling it rdma_bind_device() confusing. 

Could you modify the behavior of rdma_resolve_addr() to set the cma 
id's device field before returning? If so, that would be better than 
adding a new function.

In any event, I don't find the functionality very interesting. If the 
address resolved properly, it would speed up setup time (of course 
setup is not generally a bottleneck). In the error case (which is when 
address resolution would take a long time) things aren't any faster. 
Also consumers would still need to handle an asynchronous event.