Roland Dreier wrote:
> I assumed that since the PD protection couldn't be enforced, that
> cxgb3 shouldn't take this approach. But yea, if we ignore the pd
> issue, cxgb3 could hide this in a dma-mr, for local access only.
I guess this is probably the right idea. It's a little unfortunate that
ULPs have to try the local_lkey and then fall back to get_dma_mr() but
oh well.
Do all the current IB hca's support the local_dma_lkey? I mean, do we
really need a capability bit?
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
