[ofa-general] [PATCH RFC] RDMA/cxgb3: Fixes for zero-stag.

[Or Gerlitz]

Steve Wise wrote:
> Sort of.  It is not exactly a dma_mr.  dma_mrs allow remote IO.  The 
> zero-stag is an iWARP verbs concept.  zero-stag can be used by kernel 
> users as the lkey in any SGE that requires only local access rights.  
> If the lkey is zero, then the rnic interprets the addr field in the 
> SGE as a bus address.
>
>
I see.

Assuming we want to narrow the differences between RDMA transports from 
the ULP point of view and that IB does not support this verb - I'd like 
to see if its possible to have zero stags implemented as dma mrs.

How about enhancing the iWARP HW drivers implementation of get_dma_mr to 
switch on the access flags provided by the caller, and if no remote 
access it needed, just return zero stag for mr.lkey without doing any 
registration.


Or.