On Mon, 2007-05-07 at 10:03 -0700, Sean Hefty wrote: > > Here are a few comments. Consider them for inclusion, but what you've > > done so far is a great start. > > Thanks for the feedback. I'll try to update this before RC3 freezes. > > > - rdma_disconnect - for iWARP connections, this initiates a RDMAC Verbs > > "normal close". If the connection was properly quiesced by the > > application, then the QP will end up back in IDLE, but if the connection > > was not quiesced, then the connection will be terminated and the QP will > > end up in ERROR. Dunno if we want to describe this in detail? > > Are all work requests flushed in both cases? I don't know if we need to go > into > details about which state the QP ends up in, unless the behavior differences > are > visible to the user. >
In the "normal close" case, the user is responsible to quiesce the SQ. In both cases the RQ entries are flused. We can omit this for now if you want. > > - Also, it might be nice to have some sort of overview man page that > > maps the exected event flows for connection setup and teardown. Maybe > > 'man rdmacm' gets you some overview? > > I agree that this would be nice. Is there a standard way of doing this? > There's a 'tcp' man page to describe tcp. So I think its ok to have a 'rdmacm' or 'rdmacma' man page. Steve. _______________________________________________ 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
