From: Xiubo Li <xiubli@xxxxxxxxxx>
The reconnect feature never been supported by MDS in mds non-RECONNECT
state. This reconnect requests will incorrectly close the just reopened
sessions when the MDS kills them during the "mds_session_blocklist_on_evict"
option is disabled.
Fixes: 7e70f0ed9f3e ("ceph: attempt mds reconnect if mds closes our session")
URL: https://tracker.ceph.com/issues/65647
Signed-off-by: Xiubo Li <xiubli@xxxxxxxxxx>
---
V2:
- Try to cleanup the sessions and retry the requests.
fs/ceph/mds_client.c | 35 +++++++++++++++++++++++++++++++++--
1 file changed, 33 insertions(+), 2 deletions(-)
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index f5b25d178118..50c06a03b5fe 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -6241,9 +6241,40 @@ static void mds_peer_reset(struct ceph_connection *con)
pr_warn_client(mdsc->fsc->client, "mds%d closed our session\n",
s->s_mds);
- if (READ_ONCE(mdsc->fsc->mount_state) != CEPH_MOUNT_FENCE_IO &&
- ceph_mdsmap_get_state(mdsc->mdsmap, s->s_mds) >= CEPH_MDS_STATE_RECONNECT)
+
+ if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_FENCE_IO ||
+ ceph_mdsmap_get_state(mdsc->mdsmap, s->s_mds) < CEPH_MDS_STATE_RECONNECT)
+ return;
+
+ if (ceph_mdsmap_get_state(mdsc->mdsmap, s->s_mds) == CEPH_MDS_STATE_RECONNECT) {
send_mds_reconnect(mdsc, s);
+ return;
+ }
+
+ mutex_lock(&s->s_mutex);
+ switch (s->s_state) {
+ case CEPH_MDS_SESSION_CLOSING:
+ case CEPH_MDS_SESSION_OPEN:
+ case CEPH_MDS_SESSION_OPENING:
+ mutex_lock(&mdsc->mutex);
+ ceph_get_mds_session(s);
+ __unregister_session(mdsc, s);
+ mutex_unlock(&mdsc->mutex);
+
+ s->s_state = CEPH_MDS_SESSION_CLOSED;
+ cleanup_session_requests(mdsc, s);
+ remove_session_caps(s);
+ wake_up_all(&mdsc->session_close_wq);
+
+ mutex_lock(&mdsc->mutex);
+ __wake_requests(mdsc, &s->s_waiting);
+ kick_requests(mdsc, s->s_mds);
+ mutex_unlock(&mdsc->mutex);
+
+ ceph_put_mds_session(s);
+ break;
+ }
+ mutex_unlock(&s->s_mutex);
}
static void mds_dispatch(struct ceph_connection *con, struct ceph_msg *msg)
--
2.44.0
