Content-Type: text/plain;
        charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 11.5 \(3445.9.1\))
Subject: Re: [PATCH] lockd: NSMPROC_MON should be send only once even if in
 multithread
From:   Chuck Lever <chucklever@gmail.com>
In-Reply-To: <1547705746-69554-1-git-send-email-zhangxiaoxu5@huawei.com>
Date:   Thu, 17 Jan 2019 09:33:14 -0500
Cc:     Bruce Fields <bfields@fieldses.org>,
        Jeff Layton <jlayton@kernel.org>,
        Trond Myklebust <trond.myklebust@hammerspace.com>,
        Anna Schumaker <anna.schumaker@netapp.com>,
        Linux NFS Mailing List <linux-nfs@vger.kernel.org>
Content-Transfer-Encoding: quoted-printable
Message-Id: <4E812155-4DB4-4EF4-91ED-EDBB1B0BBFF7@gmail.com>
References: <1547705746-69554-1-git-send-email-zhangxiaoxu5@huawei.com>
To:     ZhangXiaoxu <zhangxiaoxu5@huawei.com>
Sender: linux-nfs-owner@vger.kernel.org
Precedence: bulk


> On Jan 17, 2019, at 1:15 AM, ZhangXiaoxu <zhangxiaoxu5@huawei.com> =
wrote:
>=20
> If multiple processes are locking the same file, the NSMPROC_MON
> message would be send more than one times when the host is not
> monitored by the peer.

Hi, what is the harm of sending more than one MON request for a peer?


> Add a mutex to ensure that we just send once. If some one has send
> the msg, Just waiting for the result or try again if failed.
>=20
> Signed-off-by: ZhangXiaoxu <zhangxiaoxu5@huawei.com>
> ---
> fs/lockd/mon.c              | 10 ++++++++++
> include/linux/lockd/lockd.h |  1 +
> 2 files changed, 11 insertions(+)
>=20
> diff --git a/fs/lockd/mon.c b/fs/lockd/mon.c
> index 654594e..4ba2658 100644
> --- a/fs/lockd/mon.c
> +++ b/fs/lockd/mon.c
> @@ -155,6 +155,12 @@ int nsm_monitor(const struct nlm_host *host)
> 	if (nsm->sm_monitored)
> 		return 0;
>=20
> +	mutex_lock(&nsm->mutex);
> +	if (nsm->sm_monitored) {
> +		mutex_unlock(&nsm->mutex);
> +		return 0;
> +	}
> +
> 	/*
> 	 * Choose whether to record the caller_name or IP address of
> 	 * this peer in the local rpc.statd's database.
> @@ -165,6 +171,7 @@ int nsm_monitor(const struct nlm_host *host)
> 	if (unlikely(res.status !=3D 0))
> 		status =3D -EIO;
> 	if (unlikely(status < 0)) {
> +		mutex_unlock(&nsm->mutex);
> 		pr_notice_ratelimited("lockd: cannot monitor %s\n", =
nsm->sm_name);
> 		return status;
> 	}
> @@ -174,6 +181,8 @@ int nsm_monitor(const struct nlm_host *host)
> 		nsm_local_state =3D res.state;
> 		dprintk("lockd: NSM state changed to %d\n", =
nsm_local_state);
> 	}
> +
> +	mutex_unlock(&nsm->mutex);
> 	return 0;
> }
>=20
> @@ -284,6 +293,7 @@ static struct nsm_handle *nsm_create_handle(const =
struct sockaddr *sap,
> 	memcpy(nsm_addr(new), sap, salen);
> 	new->sm_addrlen =3D salen;
> 	nsm_init_private(new);
> +	mutex_init(&new->mutex);
>=20
> 	if (rpc_ntop(nsm_addr(new), new->sm_addrbuf,
> 					sizeof(new->sm_addrbuf)) =3D=3D =
0)
> diff --git a/include/linux/lockd/lockd.h b/include/linux/lockd/lockd.h
> index b065ef4..c56069c 100644
> --- a/include/linux/lockd/lockd.h
> +++ b/include/linux/lockd/lockd.h
> @@ -89,6 +89,7 @@ struct nsm_handle {
> 	char			*sm_name;
> 	struct sockaddr_storage	sm_addr;
> 	size_t			sm_addrlen;
> +	struct mutex		mutex;  /* should be send only once even =
more threads */
> 	unsigned int		sm_monitored : 1,
> 				sm_sticky : 1;	/* don't unmonitor */
> 	struct nsm_private	sm_priv;
> --=20
> 2.7.4
>=20

--
Chuck Lever
chucklever@gmail.com