ceph: don't blindly unregister session that is in opening state

handle_cap_export() may add placeholder caps to session that is in
opening state. These caps' session pointer become wild after session get
unregistered.

The fix is not to unregister session in opening state during mds failovers,
just let client to reconnect later when mds is recovered.

Link: https://tracker.ceph.com/issues/40190
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

1 files changed, 26 insertions, 33 deletions

diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 709ac3bde86e..fcea46a54622 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -3737,42 +3737,35 @@ static void check_new_map(struct ceph_mds_client *mdsc,
 		     ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
 		     ceph_session_state_name(s->s_state));
 
-		if (i >= newmap->m_num_mds ||
-		    memcmp(ceph_mdsmap_get_addr(oldmap, i),
-			   ceph_mdsmap_get_addr(newmap, i),
-			   sizeof(struct ceph_entity_addr))) {
-			if (s->s_state == CEPH_MDS_SESSION_OPENING) {
-				/* the session never opened, just close it
-				 * out now */
-				get_session(s);
-				__unregister_session(mdsc, s);
-				__wake_requests(mdsc, &s->s_waiting);
-				ceph_put_mds_session(s);
-			} else if (i >= newmap->m_num_mds) {
-				/* force close session for stopped mds */
-				get_session(s);
-				__unregister_session(mdsc, s);
-				__wake_requests(mdsc, &s->s_waiting);
-				kick_requests(mdsc, i);
-				mutex_unlock(&mdsc->mutex);
+		if (i >= newmap->m_num_mds) {
+			/* force close session for stopped mds */
+			get_session(s);
+			__unregister_session(mdsc, s);
+			__wake_requests(mdsc, &s->s_waiting);
+			mutex_unlock(&mdsc->mutex);
 
-				mutex_lock(&s->s_mutex);
-				cleanup_session_requests(mdsc, s);
-				remove_session_caps(s);
-				mutex_unlock(&s->s_mutex);
+			mutex_lock(&s->s_mutex);
+			cleanup_session_requests(mdsc, s);
+			remove_session_caps(s);
+			mutex_unlock(&s->s_mutex);
 
-				ceph_put_mds_session(s);
+			ceph_put_mds_session(s);
 
-				mutex_lock(&mdsc->mutex);
-			} else {
-				/* just close it */
-				mutex_unlock(&mdsc->mutex);
-				mutex_lock(&s->s_mutex);
-				mutex_lock(&mdsc->mutex);
-				ceph_con_close(&s->s_con);
-				mutex_unlock(&s->s_mutex);
-				s->s_state = CEPH_MDS_SESSION_RESTARTING;
-			}
+			mutex_lock(&mdsc->mutex);
+			kick_requests(mdsc, i);
+			continue;
+		}
+
+		if (memcmp(ceph_mdsmap_get_addr(oldmap, i),
+			   ceph_mdsmap_get_addr(newmap, i),
+			   sizeof(struct ceph_entity_addr))) {
+			/* just close it */
+			mutex_unlock(&mdsc->mutex);
+			mutex_lock(&s->s_mutex);
+			mutex_lock(&mdsc->mutex);
+			ceph_con_close(&s->s_con);
+			mutex_unlock(&s->s_mutex);
+			s->s_state = CEPH_MDS_SESSION_RESTARTING;
 		} else if (oldstate == newstate) {
 			continue;  /* nothing new with this mds */
 		}