Return-Path: Received: from merit-proxy02.merit.edu ([207.75.116.194]:44044 "EHLO merit-proxy02.merit.edu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755691Ab1HYWc5 (ORCPT ); Thu, 25 Aug 2011 18:32:57 -0400 Date: Thu, 25 Aug 2011 18:32:53 -0400 From: Jim Rees To: Chuck Lever Cc: Steve Dickson , Luk Claes , linux-nfs@vger.kernel.org Subject: Re: [PATCH] mount.nfs: Preserve any explicit port=2049 option Message-ID: <20110825223253.GA7607@merit.edu> References: <1312625496-2496-1-git-send-email-luk@debian.org> <4E56AFCB.2020803@RedHat.com> Content-Type: text/plain; charset=us-ascii In-Reply-To: Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 Chuck Lever wrote: But we should be careful here: mount.nfs will do an rpcbind query if a port= was specified if there is also some doubt about whether to use TCP or UDP, or what NFS version is available. The only time rpcbind queries should be completely squashed is when the mount parameters are specified completely (transport, version, and port). Was support for SRV records ever added to either rpcbind or nfs-utils? Wouldn't that eliminate the need for portmap at least in some cases? I remember Andy working on a draft for this ages ago.