Merge tag 'ceph-for-6.2-rc7' of https://github.com/ceph/ceph-client

Pull ceph fix from Ilya Dryomov:
 "A safeguard to prevent the kernel client from further damaging the
  filesystem after running into a case of an invalid snap trace.

  The root cause of this metadata corruption is still being investigated
  but it appears to be stemming from the MDS. As such, this is the best
  we can do for now"

* tag 'ceph-for-6.2-rc7' of https://github.com/ceph/ceph-client:
  ceph: blocklist the kclient when receiving corrupted snap trace
  ceph: move mount state enum to super.h

1 files changed, 27 insertions, 3 deletions

diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 26a0a8b9975e..e163f58ff3c5 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -806,6 +806,9 @@ static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
 {
 	struct ceph_mds_session *s;
 
+	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_FENCE_IO)
+		return ERR_PTR(-EIO);
+
 	if (mds >= mdsc->mdsmap->possible_max_rank)
 		return ERR_PTR(-EINVAL);
 
@@ -1478,6 +1481,9 @@ static int __open_session(struct ceph_mds_client *mdsc,
 	int mstate;
 	int mds = session->s_mds;
 
+	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_FENCE_IO)
+		return -EIO;
+
 	/* wait for mds to go active? */
 	mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
 	dout("open_session to mds%d (%s)\n", mds,
@@ -2860,6 +2866,11 @@ static void __do_request(struct ceph_mds_client *mdsc,
 		return;
 	}
 
+	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_FENCE_IO) {
+		dout("do_request metadata corrupted\n");
+		err = -EIO;
+		goto finish;
+	}
 	if (req->r_timeout &&
 	    time_after_eq(jiffies, req->r_started + req->r_timeout)) {
 		dout("do_request timed out\n");
@@ -3245,6 +3256,7 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
 	u64 tid;
 	int err, result;
 	int mds = session->s_mds;
+	bool close_sessions = false;
 
 	if (msg->front.iov_len < sizeof(*head)) {
 		pr_err("mdsc_handle_reply got corrupt (short) reply\n");
@@ -3351,10 +3363,17 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
 	realm = NULL;
 	if (rinfo->snapblob_len) {
 		down_write(&mdsc->snap_rwsem);
-		ceph_update_snap_trace(mdsc, rinfo->snapblob,
+		err = ceph_update_snap_trace(mdsc, rinfo->snapblob,
 				rinfo->snapblob + rinfo->snapblob_len,
 				le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
 				&realm);
+		if (err) {
+			up_write(&mdsc->snap_rwsem);
+			close_sessions = true;
+			if (err == -EIO)
+				ceph_msg_dump(msg);
+			goto out_err;
+		}
 		downgrade_write(&mdsc->snap_rwsem);
 	} else {
 		down_read(&mdsc->snap_rwsem);
@@ -3412,6 +3431,10 @@ out_err:
 				     req->r_end_latency, err);
 out:
 	ceph_mdsc_put_request(req);
+
+	/* Defer closing the sessions after s_mutex lock being released */
+	if (close_sessions)
+		ceph_mdsc_close_sessions(mdsc);
 	return;
 }
 
@@ -5011,7 +5034,7 @@ static bool done_closing_sessions(struct ceph_mds_client *mdsc, int skipped)
 }
 
 /*
- * called after sb is ro.
+ * called after sb is ro or when metadata corrupted.
  */
 void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
 {
@@ -5301,7 +5324,8 @@ static void mds_peer_reset(struct ceph_connection *con)
 	struct ceph_mds_client *mdsc = s->s_mdsc;
 
 	pr_warn("mds%d closed our session\n", s->s_mds);
-	send_mds_reconnect(mdsc, s);
+	if (READ_ONCE(mdsc->fsc->mount_state) != CEPH_MOUNT_FENCE_IO)
+		send_mds_reconnect(mdsc, s);
 }
 
 static void mds_dispatch(struct ceph_connection *con, struct ceph_msg *msg)