From: Chuck Lever Subject: Re: libtirpc and nis Date: Thu, 4 Jun 2009 13:44:22 -0400 Message-ID: References: <200906031002.38551.vapier@gentoo.org> <200906031844.34398.vapier@gentoo.org> <047CD688-878B-4504-BD59-3C0B65C1B631@oracle.com> <200906032012.21876.vapier@gentoo.org> <4A279C05.9030108@RedHat.com> <1244128639.5203.44.camel@heimdal.trondhjem.org> Mime-Version: 1.0 (Apple Message framework v935.3) Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes Cc: Steve Dickson , Mike Frysinger , Linux NFS Mailing list , libtirpc To: Trond Myklebust Return-path: Received: from acsinet12.oracle.com ([141.146.126.234]:46831 "EHLO acsinet12.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752510AbZFDRoz (ORCPT ); Thu, 4 Jun 2009 13:44:55 -0400 In-Reply-To: <1244128639.5203.44.camel-rJ7iovZKK19ZJLDQqaL3InhyD016LWXt@public.gmane.org> Sender: linux-nfs-owner@vger.kernel.org List-ID: On Jun 4, 2009, at 11:17 AM, Trond Myklebust wrote: > On Thu, 2009-06-04 at 10:28 -0400, Chuck Lever wrote: >> On Jun 4, 2009, at 6:03 AM, Steve Dickson wrote: >>> Mike Frysinger wrote: >>>> On Wednesday 03 June 2009 19:59:09 Chuck Lever wrote: >>>>> On Jun 3, 2009, at 6:44 PM, Mike Frysinger wrote: >>>>>> On Wednesday 03 June 2009 11:14:44 Steve Dickson wrote: >>>>>>> Mike Frysinger wrote: >>>>>>>> the libtirpc package is self described as a "standalone >>>>>>>> package". i >>>>>>>> wonder how far that actually goes. wrt NIS, it requires the >>>>>>>> system C >>>>>>>> library to provide NIS functionality or you get a build failure >>>>>>>> in >>>>>>>> a few >>>>>>>> files. would be nice if libtirpc were usable without any NIS >>>>>>>> baggage at >>>>>>>> all. >>>>>>> NIS used RPC procedures to communicate. As it stands today >>>>>>> NIS is currently using the RPC procedures glibc. In the >>>>>>> future there is a very good chance that NIS will start >>>>>>> using the RPC procedures in libtirpc especially if >>>>>>> IPV6 support is needed... >>>>>>> >>>>>>> So I'm a bit confused on what you mean by "NIS baggage" >>>>>>> in the libtirpc package. >>>>>> you cannot build libtirpc on a system that doesnt provide NIS >>>>>> functionality. >>>>>> realistic RPC usage today is NFS related services only (i.e. a >>>>>> NFS >>>>>> client >>>>>> mounting a share on a NFS server). >>>>> Mike, it would help if you could provide the build output so we >>>>> can >>>>> see exactly what's failing. >>>> >>>> libtirpc does not provide rpcsvc/nis.h >>> >>> Hmm... I wonder what it would take to incorporate that file into >>> libitrpc... >> >> Alternately, you could disable auth_des on systems that don't already >> have NIS. > > Why? It isn't using any NIS functionality here. All that is being done > is to store information in a structure. src/auth_time.c appears to assume IPv4. It handles its own universal address translation, duplicating code that can be found in other parts of the library. The AUTH_DES code might call rpcb_gettime() instead, and simply drop auth_time.c, but the RPCB_GETTIME call won't work against our legacy portmapper. It looks like no-one has visited this code in quite some time. In addition, my impression was that AUTH_DES, the only caller of __rpc_get_time_offset, has been deprecated in favor of AUTH_GSS. I don't think Linux kernel RPC supports AUTH_DES, for instance. So, why spend any effort trying to make this work? I'm suggesting we can shit-can AUTH_DES support in libtirpc at least if NIS headers aren't present on the system, and perhaps even for all configurations. -- Chuck Lever chuck[dot]lever[at]oracle[dot]com