Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 23.128.96.31 as permitted sender) client-ip=23.128.96.31;
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
From:   "NeilBrown" <neilb@suse.de>
To:     trondmy@kernel.org
Cc:     "Anna Schumaker" <Anna.Schumaker@netapp.com>,
        linux-nfs@vger.kernel.org
Subject: Re: [PATCH 1/2] NFSv4: Fix a nfs4_state_manager() race
In-reply-to: <20230917230551.30483-1-trondmy@kernel.org>
References: <20230917230551.30483-1-trondmy@kernel.org>
Date:   Mon, 18 Sep 2023 11:17:35 +1000
Message-id: <169499985553.8274.16707412085976640956@noble.neil.brown.name>
Precedence: bulk

On Mon, 18 Sep 2023, trondmy@kernel.org wrote:
> From: Trond Myklebust <trond.myklebust@hammerspace.com>
>=20
> If the NFS4CLNT_RUN_MANAGER flag got set just before we cleared
> NFS4CLNT_MANAGER_RUNNING, then we might have won the race against
> nfs4_schedule_state_manager(), and are responsible for handling the
> recovery situation.
>=20
> Fixes: aeabb3c96186 ("NFSv4: Fix a NFSv4 state manager deadlock")
> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
> ---
>  fs/nfs/nfs4state.c | 7 +++++++
>  1 file changed, 7 insertions(+)
>=20
> diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
> index e079987af4a3..0bc160fbabec 100644
> --- a/fs/nfs/nfs4state.c
> +++ b/fs/nfs/nfs4state.c
> @@ -2703,6 +2703,13 @@ static void nfs4_state_manager(struct nfs_client *cl=
p)
>  		nfs4_end_drain_session(clp);
>  		nfs4_clear_state_manager_bit(clp);
> =20
> +		if (test_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state) &&
> +		    !test_and_set_bit(NFS4CLNT_MANAGER_RUNNING,
> +				      &clp->cl_state)) {
> +			memflags =3D memalloc_nofs_save();
> +			continue;
> +		}
> +

I cannot see what race this closes.
When we cleared MANAGER_RUNNING, the only possible consequence is that
nfs4_wait_clnt_recover could have woken up.  This leads to
nfs4_schedule_state_manager() being run, which sets RUN_MANAGER whether
it was set or not.

I understand that there are problems with MANAGER_AVAILABLE which your
next patch addresses, but I cannot see what this one actually fixes.
Can you help me?

Thanks,
NeilBrown


>  		if (!test_and_set_bit(NFS4CLNT_RECALL_RUNNING, &clp->cl_state)) {
>  			if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
>  				nfs_client_return_marked_delegations(clp);
> --=20
> 2.41.0
>=20
>=20