Subject: Re: [PATCH 4/4] autofs4 - add miscelaneous device for ioctls
From: Ian Kent <raven@themaw.net>
To: Christoph Hellwig <hch@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>,
       autofs mailing list <autofs@linux.kernel.org>,
       Kernel Mailing List <linux-kernel@vger.kernel.org>,
       linux-fsdevel <linux-fsdevel@vger.kernel.org>
In-Reply-To: <20080809171816.GA2039@infradead.org>
References: <20080807114002.4142.30417.stgit@web.messagingengine.com>
	 <20080807114030.4142.76568.stgit@web.messagingengine.com>
	 <20080809125953.GB16308@infradead.org>
	 <1218295750.2994.61.camel@raven.themaw.net>
	 <20080809171816.GA2039@infradead.org>
Content-Type: text/plain
Date: Sun, 10 Aug 2008 13:20:37 +0800
Message-Id: <1218345638.2979.6.camel@raven.themaw.net>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2493
Lines: 49


On Sat, 2008-08-09 at 13:18 -0400, Christoph Hellwig wrote:
> On Sat, Aug 09, 2008 at 11:29:09PM +0800, Ian Kent wrote:
> > 
> > On Sat, 2008-08-09 at 08:59 -0400, Christoph Hellwig wrote:
> > > On Thu, Aug 07, 2008 at 07:40:31PM +0800, Ian Kent wrote:
> > > > The actual problem with autofs is that it can't reconnect to existing
> > > > mounts. Immediately one things of just adding the ability to remount
> > > > autofs file systems would solve it, but alas, that can't work. This is
> > > > because autofs direct mounts and the implementation of "on demand mount
> > > > and expire" of nested mount trees have the file system mounted on top of
> > > > the mount trigger dentry.
> > > 
> > > So what you really need instead of all the ioctl cruft is to get access
> > > to the sb of the hidden autofs4 mount.  One way to do that that I can
> > > think of right now is to change from using get_sb_nodev as ->get_sb
> > > to a variant that can find an existing superblock using some mount
> > > options.
> > 
> > I wished I'd spoken to you about this a long time ago since you seem to
> > know how I can do this using a remount option but this isn't enough of
> > an explanation of how it can be done.
> > 
> > AFAICT, mount(8) will never be able to get a "struct path" or "struct
> > nameidata" to a mount point that is covered using the path lookup in
> > fs/namespace.c:do_*_mount() so we'll never be called back via any VFS
> > methods. So tell me more about how this could work please!
> 
> Your introduction tells what you want to do is a remount, for which
> finding the superblock is enough.  If that's not the actual requirement
> please explain the requirement.  And don't point me to the gazillion
> ioctls you added, that seems to mostly not have much to do with that..

The text you've quoted from provides quite a detailed explanation of
this and my response above to your question also notes why this cannot
be done in this case.

Perhaps you missed the fact that the mount in question is covered by
another mount, as noted in the quoted text and in my answer, which is
why a mount with a remount option cannot work. Or perhaps your assuming
the covering mount is also an autofs mount which may be the case but
generally is not.


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/