[openib-general][RFC]: CMA IB implementation

[Guy German]

Sean Hefty wrote:
> I think that there will still be a need for a separate address 
> translation module(s)

I don't understand. You think there should be an address translation module, but 
you object to the *name* ib_at ? (ib_at stands for "infiniband address translation")

I suggested before that if ib_at should be fixed lets fix it. If API should be 
improved or other functionality should be added (or removed) why not do it in 
the existing ib_at ?

If there will indeed be a separate address translation module((s)?), then why 
would transport aware modules won't use it along with the cm ?
Doesn't that leave the cma with the abstraction purposes only ?

> Caching will be complex, which is why I think that it needs to have its 
> own module.  I'm envisioning a cache that can be saved to disk for 
> faster system startup.

I admit I'm not aware to all the complexity of caching, so I fail to see why it 
can't be implemented in the ib_at module.


The way I see it - If the cma can replace ib_at functionality and also serve as 
an at module, that's fine. But if we decide that there should be an ib_at module 
(which centralizes the at to all the ULP's) the cma should use this module and 
the cma consumer needn't be aware of it.

Guy