Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <d55ffcb38edc125a8b91c6a885e01c857238caf9.camel@kernel.org>
Subject: Re: [RFC PATCH 0/3] nfsd: close potential race between open and
 setting delegation
From:   Jeff Layton <jlayton@kernel.org>
To:     "J. Bruce Fields" <bfields@fieldses.org>
Cc:     chuck.lever@oracle.com, neilb@suse.de, linux-nfs@vger.kernel.org
Date:   Mon, 18 Jul 2022 17:43:06 -0400
In-Reply-To: <20220718211819.GA28925@fieldses.org>
References: <20220714152819.128276-1-jlayton@kernel.org>
         <20220718211819.GA28925@fieldses.org>
Content-Type: text/plain; charset="ISO-8859-15"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.44.3 (3.44.3-1.fc36) 
MIME-Version: 1.0
Precedence: bulk

On Mon, 2022-07-18 at 17:18 -0400, J. Bruce Fields wrote:
> On Thu, Jul 14, 2022 at 11:28:16AM -0400, Jeff Layton wrote:
> > Here's a first stab at a patchset to close a potential race when settin=
g
> > a delegation on a file. Between the point where we open the file and
> > where we set the delegation, another task or client could unlink or
> > rename the dentry. If that occurs, we shouldn't hand out a delegation
> > in the open response, but we don't prevent that today.
> >=20
> > The basic idea here is to re-do the lookup after setting the delegation=
.
> > If the resulting dentry is not the one we have in the open, then we can
> > reject handing out a delegation.
>=20
> I have this distinct memory of actually doing that before.
>=20
> But looking through the git history all I find is 4335723e8e9f "nfsd4:
> fix delegation-unlink/rename race", from 2014, which claims to fix a
> similar-sounding race in a different way.
>=20
> How are you reproducing this?
>=20
> --b.
>=20

I'm not at all, so far. This race is entirely theoretical. Probably we
could reproduce it by introducing some artificial delays or something.

> >=20
> > Only lightly tested, so this is an RFC for now.
> >=20
> > Jeff Layton (3): nfsd: drop fh argument from alloc_init_deleg nfsd:
> > rework arguments to nfs4_set_delegation nfsd: vet the opened dentry
> > after setting a delegation
> >=20
> >  fs/nfsd/nfs4state.c | 65
> >  ++++++++++++++++++++++++++++++++++++++------- 1 file changed, 55
> >  insertions(+), 10 deletions(-)
> >=20
> > -- 2.36.1

--=20
Jeff Layton <jlayton@kernel.org>