From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from cantor2.suse.de ([195.135.220.15]:47533 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755958Ab1F1CLz (ORCPT ); Mon, 27 Jun 2011 22:11:55 -0400 Date: Tue, 28 Jun 2011 12:11:44 +1000 From: NeilBrown To: Steve Dickson Cc: Chuck Lever , Linux NFS Mailing List Subject: Re: [PATCH/RFC] mount: improve signal management when locking mtab. Message-ID: <20110628121144.0afb228e@notabene.brown> In-Reply-To: <4E08B7D8.6050405@RedHat.com> References: <20110622163852.71a2ae80@notabene.brown> <20110623092052.414d78a8@notabene.brown> <4D5CF8CE-FDFE-491D-99A4-D35853DDEA61@oracle.com> <4E08B7D8.6050405@RedHat.com> Content-Type: text/plain; charset=US-ASCII Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 On Mon, 27 Jun 2011 13:03:20 -0400 Steve Dickson wrote: > > > On 06/23/2011 12:09 PM, Chuck Lever wrote: > > > > On Jun 22, 2011, at 5:20 PM, NeilBrown wrote: > > > >> On Wed, 22 Jun 2011 09:08:25 -0600 Chuck Lever wrote: > >> > >>> Since we are going to adopt libmount for mount.nfs in the near future, would it be better to update libmount instead? > >> > >> How near? > >> > >> Libmount appears to get signal handling right already (though I haven't > >> tested it). It blocks all signals rather than catching some of them. > >> > >> So switching to libmount would a perfectly reasonably response. However if > >> that is more than a few weeks away I think I would rather see this fixed up > >> anyway... > > > > We have a libmount-based mount.nfs already in the nfs-utils tree, IIRC. I don't think we yet have a generic plan for switching to installing that one by default. Steve? > Well a while back I did add the libmount code along with --enable-libmount-mount > configure flag. This flag has been enabled for a while now in the > pre release of Fedora 16 so it will on in the release of f16. > > Plus I am looking to make a nfs-utils release... I'm in the process of > clean things up just to do that. So I would be willing to enable the libmount > code to be on by default. But I am concern not all distros do include > the libmount code... > > Some clarity... If the libmount code is enabled, this patch is not needed? > No it is not ... however it would be nice if the affected code wasn't even compiled in. i.e. which a patch like this, it will not even compile the offending code if libmount is selected. NeilBrown diff --git a/utils/mount/Makefile.am b/utils/mount/Makefile.am index 056293c..7bc3e2b 100644 --- a/utils/mount/Makefile.am +++ b/utils/mount/Makefile.am @@ -9,11 +9,10 @@ man5_MANS = nfs.man sbin_PROGRAMS = mount.nfs EXTRA_DIST = nfsmount.x $(man8_MANS) $(man5_MANS) -mount_common = error.c network.c fstab.c token.c \ +mount_common = error.c network.c token.c \ parse_opt.c parse_dev.c \ nfsmount.c nfs4mount.c stropts.c\ - nfsumount.c \ - mount_constants.h error.h network.h fstab.h token.h \ + mount_constants.h error.h network.h token.h \ parse_opt.h parse_dev.h \ nfs4_mount.h nfs_mount4.h stropts.h version.h \ mount_config.h utils.c utils.h @@ -33,7 +32,8 @@ if CONFIG_LIBMOUNT mount_nfs_SOURCES += mount_libmount.c mount_nfs_LDADD += $(LIBMOUNT) else -mount_nfs_SOURCES += mount.c +mount_nfs_SOURCES += mount.c fstab.c nfsumount.c fstab.h + endif MAINTAINERCLEANFILES = Makefile.in