Re: [PATCH 2/3] ceph: don't WARN if we're force umounting

From: Jeff Layton <jlayton@kernel.org>
To: xiubli@redhat.com
Cc: idryomov@gmail.com, pdonnell@redhat.com, ceph-devel@vger.kernel.org
Subject: Re: [PATCH 2/3] ceph: don't WARN if we're force umounting
Date: Mon, 23 Aug 2021 09:49:17 -0400	[thread overview]
Message-ID: <7bf49c80528b31f6350d7f3ee2a5a69da42aaa69.camel@kernel.org> (raw)
In-Reply-To: <20210818080603.195722-3-xiubli@redhat.com>

On Wed, 2021-08-18 at 16:06 +0800, xiubli@redhat.com wrote:
> From: Xiubo Li <xiubli@redhat.com>
> 
> Force umount will try to close the sessions by setting the session
> state to _CLOSING, so in ceph_kill_sb after that it will warn on it.
> 
> URL: https://tracker.ceph.com/issues/52295
> Signed-off-by: Xiubo Li <xiubli@redhat.com>
> ---
>  fs/ceph/mds_client.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> index a632e1c7cef2..0302af53e079 100644
> --- a/fs/ceph/mds_client.c
> +++ b/fs/ceph/mds_client.c
> @@ -4558,6 +4558,8 @@ static void maybe_recover_session(struct ceph_mds_client *mdsc)
>  
>  bool check_session_state(struct ceph_mds_session *s)
>  {
> +	struct ceph_fs_client *fsc = s->s_mdsc->fsc;
> +
>  	switch (s->s_state) {
>  	case CEPH_MDS_SESSION_OPEN:
>  		if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
> @@ -4566,8 +4568,11 @@ bool check_session_state(struct ceph_mds_session *s)
>  		}
>  		break;
>  	case CEPH_MDS_SESSION_CLOSING:
> -		/* Should never reach this when we're unmounting */
> -		WARN_ON_ONCE(s->s_ttl);
> +		/*
> +		 * Should never reach this when none force unmounting
> +		 */
> +		if (READ_ONCE(fsc->mount_state) != CEPH_MOUNT_SHUTDOWN)
> +			WARN_ON_ONCE(s->s_ttl);

How about something like this instead?

    WARN_ON_ONCE(s->s_ttl && READ_ONCE(fsc->mount_state) != CEPH_MOUNT_SHUTDOWN);

>  		fallthrough;
>  	case CEPH_MDS_SESSION_NEW:
>  	case CEPH_MDS_SESSION_RESTARTING:

-- 
Jeff Layton <jlayton@kernel.org>