ceph: auto reconnect after blacklisted

Make client use osd reply and session message to infer if itself is
blacklisted. Client reconnect to cluster using new entity addr if it
is blacklisted. Auto reconnect is limited to once every 30 minutes.

Auto reconnect is disabled by default. It can be enabled/disabled by
recover_session=<no|clean> mount option. In 'clean' mode, client drops
any dirty data/metadata, invalidates page caches and invalidates all
writable file handles. After reconnect, file locks become stale because
MDS loses track of them. If an inode contains any stale file locks,
read/write on the indoe are not allowed until applications release all
stale file locks.

Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

1 files changed, 32 insertions, 2 deletions

diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index ed4d20a54af7..5bfbff8f2f64 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -3032,18 +3032,23 @@ bad:
 	pr_err("mdsc_handle_forward decode error err=%d\n", err);
 }
 
-static int __decode_and_drop_session_metadata(void **p, void *end)
+static int __decode_session_metadata(void **p, void *end,
+				     bool *blacklisted)
 {
 	/* map<string,string> */
 	u32 n;
+	bool err_str;
 	ceph_decode_32_safe(p, end, n, bad);
 	while (n-- > 0) {
 		u32 len;
 		ceph_decode_32_safe(p, end, len, bad);
 		ceph_decode_need(p, end, len, bad);
+		err_str = !strncmp(*p, "error_string", len);
 		*p += len;
 		ceph_decode_32_safe(p, end, len, bad);
 		ceph_decode_need(p, end, len, bad);
+		if (err_str && strnstr(*p, "blacklisted", len))
+			*blacklisted = true;
 		*p += len;
 	}
 	return 0;
@@ -3067,6 +3072,7 @@ static void handle_session(struct ceph_mds_session *session,
 	u64 seq;
 	unsigned long features = 0;
 	int wake = 0;
+	bool blacklisted = false;
 
 	/* decode */
 	ceph_decode_need(&p, end, sizeof(*h), bad);
@@ -3079,7 +3085,7 @@ static void handle_session(struct ceph_mds_session *session,
 	if (msg_version >= 3) {
 		u32 len;
 		/* version >= 2, metadata */
-		if (__decode_and_drop_session_metadata(&p, end) < 0)
+		if (__decode_session_metadata(&p, end, &blacklisted) < 0)
 			goto bad;
 		/* version >= 3, feature bits */
 		ceph_decode_32_safe(&p, end, len, bad);
@@ -3166,6 +3172,8 @@ static void handle_session(struct ceph_mds_session *session,
 		session->s_state = CEPH_MDS_SESSION_REJECTED;
 		cleanup_session_requests(mdsc, session);
 		remove_session_caps(session);
+		if (blacklisted)
+			mdsc->fsc->blacklisted = true;
 		wake = 2; /* for good measure */
 		break;
 
@@ -4015,7 +4023,27 @@ static void lock_unlock_sessions(struct ceph_mds_client *mdsc)
 	mutex_unlock(&mdsc->mutex);
 }
 
+static void maybe_recover_session(struct ceph_mds_client *mdsc)
+{
+	struct ceph_fs_client *fsc = mdsc->fsc;
+
+	if (!ceph_test_mount_opt(fsc, CLEANRECOVER))
+		return;
+
+	if (READ_ONCE(fsc->mount_state) != CEPH_MOUNT_MOUNTED)
+		return;
 
+	if (!READ_ONCE(fsc->blacklisted))
+		return;
+
+	if (fsc->last_auto_reconnect &&
+	    time_before(jiffies, fsc->last_auto_reconnect + HZ * 60 * 30))
+		return;
+
+	pr_info("auto reconnect after blacklisted\n");
+	fsc->last_auto_reconnect = jiffies;
+	ceph_force_reconnect(fsc->sb);
+}
 
 /*
  * delayed work -- periodically trim expired leases, renew caps with mds
@@ -4089,6 +4117,8 @@ static void delayed_work(struct work_struct *work)
 
 	ceph_trim_snapid_map(mdsc);
 
+	maybe_recover_session(mdsc);
+
 	schedule_delayed(mdsc);
 }