> Combining possibilities: if you want to get a notification on 
> every peer process crash, you would need to either 
> poll/select once a while the libibcm/librdmacm event queue or 
> implement some keep a live of your own protocol. For 
> instance, I think the IB spec mentions doing zero length rdma 
> write once in a while as a mean for implementing such protocol.


Can you point me the spec page talking about zero-length rdma-write and
send ?

If I use zero-length rdma write, does it generate something on wire to
let me detect broken connection, and there is no effect on remote buffer
?

If I use zero-length send, do I get the same thing? How about if I don't
have receive posted on remote side ?

--CQ


> 
> Or.
> 

_______________________________________________
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

Reply via email to