Date: Mon, 23 Apr 2012 11:16:10 -0400
From: Jeff Layton <jlayton@redhat.com>
To: Miklos Szeredi <miklos@szeredi.hu>
Cc: "J. Bruce Fields" <bfields@fieldses.org>,
        Malahal Naineni <malahal@us.ibm.com>,
        Steve Dickson <SteveD@redhat.com>, linux-fsdevel@vger.kernel.org,
        linux-nfs@vger.kernel.org, linux-kernel@vger.kernel.org,
        viro@zeniv.linux.org.uk, hch@infradead.org,
        michael.brantley@deshaw.com, sven.breuner@itwm.fraunhofer.de,
        chuck.lever@oracle.com, pstaubach@exagrid.com,
        trond.myklebust@fys.uio.no, rees@umich.edu
Subject: Re: [PATCH RFC v3] vfs: make fstatat retry once on ESTALE errors
 from getattr call
Message-ID: <20120423111610.0259d610@tlielax.poochiereds.net>
In-Reply-To: <87hawasdrb.fsf@tucsk.pomaz.szeredi.hu>
References: <20120420104055.511e15bc@tlielax.poochiereds.net>
	<4F91C49D.8070908@RedHat.com>
	<20120420203725.GA3512@us.ibm.com>
	<20120420171314.73801874@corrin.poochiereds.net>
	<CAJfpegt40cgMJQQo3JuNaaS1w957Y2a_NxVoyvx3bmTMj1TGOA@mail.gmail.com>
	<20120423080012.7c23ef24@tlielax.poochiereds.net>
	<20120423130009.GA13681@fieldses.org>
	<20120423091255.00f926c4@tlielax.poochiereds.net>
	<20120423133412.GB13681@fieldses.org>
	<20120423095021.1a91a23b@tlielax.poochiereds.net>
	<20120423135456.GC13681@fieldses.org>
	<87hawasdrb.fsf@tucsk.pomaz.szeredi.hu>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Sender: linux-nfs-owner@vger.kernel.org

On Mon, 23 Apr 2012 16:51:04 +0200
Miklos Szeredi <miklos@szeredi.hu> wrote:

> "J. Bruce Fields" <bfields@fieldses.org> writes:
> 
> >
> > I also wonder whether it would be making too many assumptions about the
> > server or filesystem: just because ordinary posix interfaces don't allow
> > atomic replacement of a whole directory tree doesn't mean the server
> > might not have some way to do it.
> 
> Exactly because posix limits the atomic replacement to empty directories
> is that this feature is not useful and is why linux can get away with
> the dead directory behavior in this case.  And thinking about fixing
> this in NFS is completely pointless since no one will rely on the atomic
> replacement behavior.  Fixing local filesystems is also pointless for
> the same reason.
> 
> Atomic replacement of whole directory trees would indeed be more useful,
> but it's highly unlikely to be used anywhere since applications relying
> on this feature would be limited to special filesystems that allow this.
> 
> So my statement is "ENOENT is equivalent to ESTALE if already retrying
> path lookup with LOOKUP_REVAL on any operation that takes an parent
> directory and a name (lookup, create, link, unlink, symlink, mkdir,
> rmdir, mknod, rename)."
> 

Ok, but again, that only applies to the lookup. It has no bearing on
the subsequent operation. For instance, if we're doing:

     rename("/foo", "/bar");

...and another client is simultaneously doing:

     creat("/bar/baz", 0600);

...and we get back ESTALE from the server on the create because the
"old" /bar got replaced after the lookup of it. Then it seems like
returning -ENOENT would not be correct since there was never a time
where /bar didn't exist...

It might eventually be nice to add that optimization to the path lookup
code. OTOH, it only solves a very specific problem that's not really
applicable in a lot of the cases I'm interested in fixing.

-- 
Jeff Layton <jlayton@redhat.com>