On Tue, Oct 06, 2009 at 06:59:47AM -0700, Erik Nordmark wrote: > Anders Persson wrote: >> * 9629-30: If there were any credentials, then they would have already been >> attached to the t_conn_ind mp in tcp_conn_request. So is it necessary to >> do this? > > tcp_input_data also calls it; but are you saying that tcp_eager_conn_ind > always has the db_credp set? Sure looks like it.
Yes. >> * 21407: Looks like the endpoint is not being removed from the bind >> hash > > You mean calling tcp_bind_hash_remove? No need to remove from the ipcl > one since we didn't get inserted. Yes, since tcp_bind_check() ends up inserting the conn in the tcp bind hash, tcp_bind_hash_remove() should be called in case the fanout insert fails. That was previously handled by tcp_post_ip_bind(). Cheers, Anders _______________________________________________ networking-discuss mailing list [email protected]
