Return-Path: Received: from rcsinet10.oracle.com ([148.87.113.121]:53322 "EHLO rcsinet10.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752138Ab1BRVsv convert rfc822-to-8bit (ORCPT ); Fri, 18 Feb 2011 16:48:51 -0500 Subject: Re: umount.nfs4 is disturbed by /proc/mounts Content-Type: text/plain; charset=us-ascii From: Chuck Lever In-Reply-To: <20110218195605.GD25634@nb.net.home> Date: Fri, 18 Feb 2011 13:48:31 -0800 Cc: Jeff Layton , Jan Engelhardt , linux-nfs@vger.kernel.org Message-Id: References: <20110218142828.0a35e9d7@corrin.poochiereds.net> <20110218195605.GD25634@nb.net.home> To: Karel Zak Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 On Feb 18, 2011, at 11:56 AM, Karel Zak wrote: > On Fri, Feb 18, 2011 at 02:28:28PM -0500, Jeff Layton wrote: >> On Fri, 18 Feb 2011 09:11:54 -0800 >> Chuck Lever wrote: >> >>> >>> On Feb 17, 2011, at 3:44 PM, Jan Engelhardt wrote: >>> >>>> >>>> Recent distributions start to have /etc/mtab being a symlink to >>>> /proc/mounts with util-linux 2.19. Subsequently, this annoys umount.nfs4 >>>> from nfs-client-1.2.3 and emits a warning: >>>> >>>> jng-0:/ # grep nfs4 /proc/mounts >>>> 10.10.1.1:/home/ /home nfs4 rw,relatime,vers=4,rsize=1048576,wsize=1048576,namlen=255,hard,proto=tcp,port=0,timeo=600,retrans=2,sec=sys,clientaddr=10.10.1.212,minorversion=0,local_lock=none,addr=10.10.1.1 0 0 >>>> jng-0:/ # umount /home >>>> umount.nfs4: invalid value for 'port=' option >>> >>> I can take a look at this. >>> >> >> FWIW, I just fixed mount.cifs recently based on recommendations from >> Karel: >> >> http://git.samba.org/?p=cifs-utils.git;a=commitdiff;h=fba28cfe2f13dd8bdae3cec76178f42b001a40ca >> >> ...this is the same approach that /bin/mount uses, so you're probably >> safe to use a similar one for mount/umount.nfs. > > This is not the same problem. The NFS umount/remount depends on > (userspace) values from /etc/mtab. Right, this problem is not dependent on the existence of mtab. There is a real bug here, I suggest. port=0 is a valid specification, and the mount option parser in mount.nfs and umount.nfs should accept it. That's not dependent on whether /etc/mtab is a link or whether we are using libmount to do the work. I have a very simple patch to fix this now, just needs testing. > The solution is to link mount.nfs with libmount. I'm working on this > change... http://www.spinics.net/lists/linux-nfs/msg19165.html > > > I have already talked about mtab issues at fedora-devel list, see > http://lists.fedoraproject.org/pipermail/devel/2011-January/147995.html > > The mount.nfs and pam_mount are on my TODO list for the next weeks. > If the changes in mount.nfs will be successful then I'd like to link > mount.cifs with libmount too. > > Note that mount(8) in Fedora still supports regular mtab too, and > the regular mtab is still util-linux upstream default. > > Karel > > -- > Karel Zak > http://karelzak.blogspot.com -- Chuck Lever chuck[dot]lever[at]oracle[dot]com