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: <57dc06d57b4b643b4bf04daf28acca202c9f7a85.camel@kernel.org>
Subject: Re: [PATCH 1/1] NFSD: fix WARN_ON_ONCE in __queue_delayed_work
From:   Jeff Layton <jlayton@kernel.org>
To:     Dai Ngo <dai.ngo@oracle.com>, chuck.lever@oracle.com
Cc:     efault@gmx.de, linux-nfs@vger.kernel.org
Date:   Tue, 10 Jan 2023 05:30:38 -0500
In-Reply-To: <1673333310-24837-1-git-send-email-dai.ngo@oracle.com>
References: <1673333310-24837-1-git-send-email-dai.ngo@oracle.com>
Content-Type: text/plain; charset="ISO-8859-15"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.46.2 (3.46.2-1.fc37) 
MIME-Version: 1.0
Precedence: bulk

On Mon, 2023-01-09 at 22:48 -0800, Dai Ngo wrote:
> Currently nfsd4_state_shrinker_worker can be schduled multiple times
> from nfsd4_state_shrinker_count when memory is low. This causes
> the WARN_ON_ONCE in __queue_delayed_work to trigger.
>=20
> This patch allows only one instance of nfsd4_state_shrinker_worker
> at a time using the nfsd_shrinker_active flag, protected by the
> client_lock.
>=20
> Replace mod_delayed_work with queue_delayed_work since we
> don't expect to modify the delay of any pending work.
>=20
> Fixes: 44df6f439a17 ("NFSD: add delegation reaper to react to low memory =
condition")
> Reported-by: Mike Galbraith <efault@gmx.de>
> Signed-off-by: Dai Ngo <dai.ngo@oracle.com>
> ---
>  fs/nfsd/netns.h     |  1 +
>  fs/nfsd/nfs4state.c | 16 ++++++++++++++--
>  2 files changed, 15 insertions(+), 2 deletions(-)
>=20
> diff --git a/fs/nfsd/netns.h b/fs/nfsd/netns.h
> index 8c854ba3285b..801d70926442 100644
> --- a/fs/nfsd/netns.h
> +++ b/fs/nfsd/netns.h
> @@ -196,6 +196,7 @@ struct nfsd_net {
>  	atomic_t		nfsd_courtesy_clients;
>  	struct shrinker		nfsd_client_shrinker;
>  	struct delayed_work	nfsd_shrinker_work;
> +	bool			nfsd_shrinker_active;
>  };
> =20
>  /* Simple check to find out if a given net was properly initialized */
> diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> index ee56c9466304..e00551af6a11 100644
> --- a/fs/nfsd/nfs4state.c
> +++ b/fs/nfsd/nfs4state.c
> @@ -4407,11 +4407,20 @@ nfsd4_state_shrinker_count(struct shrinker *shrin=
k, struct shrink_control *sc)
>  	struct nfsd_net *nn =3D container_of(shrink,
>  			struct nfsd_net, nfsd_client_shrinker);
> =20
> +	spin_lock(&nn->client_lock);
> +	if (nn->nfsd_shrinker_active) {
> +		spin_unlock(&nn->client_lock);
> +		return 0;
> +	}

Is this extra machinery really necessary? The bool and spinlock don't
seem to be needed. Typically there is no issue with calling
queued_delayed_work when the work is already queued. It just returns
false in that case without doing anything.

>=20
>  	count =3D atomic_read(&nn->nfsd_courtesy_clients);
>  	if (!count)
>  		count =3D atomic_long_read(&num_delegations);
> -	if (count)
> -		mod_delayed_work(laundry_wq, &nn->nfsd_shrinker_work, 0);
> +	if (count) {
> +		nn->nfsd_shrinker_active =3D true;
> +		spin_unlock(&nn->client_lock);
> +		queue_delayed_work(laundry_wq, &nn->nfsd_shrinker_work, 0);
> +	} else
> +		spin_unlock(&nn->client_lock);
>  	return (unsigned long)count;
>  }
> =20
> @@ -6239,6 +6248,9 @@ nfsd4_state_shrinker_worker(struct work_struct *wor=
k)
> =20
>  	courtesy_client_reaper(nn);
>  	deleg_reaper(nn);
> +	spin_lock(&nn->client_lock);
> +	nn->nfsd_shrinker_active =3D 0;
> +	spin_unlock(&nn->client_lock);
>  }
> =20
>  static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stid =
*stp)

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