[ofa-general] Re: consumer data buffer ownership for inline sends

[Or Gerlitz]

Roland Dreier wrote:
>  > Does this means that for inline sends, when ibv_post_send returns,
>  > the consumer owns back the data buffer associated with this send?
>  > 
>  > Can this be stated as the official policy of libibverbs?
> 
> I guess that makes sense.  I wonder if there's any conceivable
> interpretation of the inline send flag where the adapter might need to
> access the original buffer after the request is posted?

thinking on it a little, such adapter has too much logic/state 
implemented in its HW/DMA engine... assuming all this is beyond the IB 
spec scope, can we take the liberty and turn it into official policy of 
libibverbs which provider libraries must confirm to?

Or.