drbd: Ignore the exit code of a fence-peer handler if it returns too late

In case the connection was established and lost again before
the a fence-peer handler returns, ignore the exit code of this
instance. (And use the exit code of the later started instance)

Signed-off-by: Andreas Gruenbacher <agruen@linbit.com>
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

3 files changed, 17 insertions, 3 deletions

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index f943aacfdad8..f104328d6325 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -832,6 +832,7 @@ struct drbd_tconn {			/* is a resource from the config file */
         unsigned susp_nod:1;            /* IO suspended because no data */
         unsigned susp_fen:1;            /* IO suspended because fence peer handler runs */
         struct mutex cstate_mutex;      /* Protects graceful disconnects */
+        unsigned int connect_cnt;       /* Inc each time a connection is established */
 
         unsigned long flags;
         struct net_conf *net_conf;      /* content protected by rcu */
diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c
index 0936d6aabef9..e25803b447ff 100644
--- a/drivers/block/drbd/drbd_nl.c
+++ b/drivers/block/drbd/drbd_nl.c
@@ -417,6 +417,7 @@ static enum drbd_fencing_p highest_fencing_policy(struct drbd_tconn *tconn)
 
 bool conn_try_outdate_peer(struct drbd_tconn *tconn)
 {
+        unsigned int connect_cnt;
         union drbd_state mask = { };
         union drbd_state val = { };
         enum drbd_fencing_p fp;
@@ -428,6 +429,10 @@ bool conn_try_outdate_peer(struct drbd_tconn *tconn)
                 return false;
         }
 
+        spin_lock_irq(&tconn->req_lock);
+        connect_cnt = tconn->connect_cnt;
+        spin_unlock_irq(&tconn->req_lock);
+
         fp = highest_fencing_policy(tconn);
         switch (fp) {
         case FP_NOT_AVAIL:
@@ -492,8 +497,14 @@ bool conn_try_outdate_peer(struct drbd_tconn *tconn)
            here, because we might were able to re-establish the connection in the
            meantime. */
         spin_lock_irq(&tconn->req_lock);
-        if (tconn->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &tconn->flags))
-                _conn_request_state(tconn, mask, val, CS_VERBOSE);
+        if (tconn->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &tconn->flags)) {
+                if (tconn->connect_cnt != connect_cnt)
+                        /* In case the connection was established and droped
+                           while the fence-peer handler was running, ignore it */
+                        conn_info(tconn, "Ignoring fence-peer exit code\n");
+                else
+                        _conn_request_state(tconn, mask, val, CS_VERBOSE);
+        }
         spin_unlock_irq(&tconn->req_lock);
 
         return conn_highest_pdsk(tconn) <= D_OUTDATED;
diff --git a/drivers/block/drbd/drbd_state.c b/drivers/block/drbd/drbd_state.c
index 90c5be2b1d30..216d47b7e88b 100644
--- a/drivers/block/drbd/drbd_state.c
+++ b/drivers/block/drbd/drbd_state.c
@@ -1115,8 +1115,10 @@ __drbd_set_state(struct drbd_conf *mdev, union drbd_state ns,
                 drbd_thread_restart_nowait(&mdev->tconn->receiver);
 
         /* Resume AL writing if we get a connection */
-        if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED)
+        if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) {
                 drbd_resume_al(mdev);
+                mdev->tconn->connect_cnt++;
+        }
 
         /* remember last attach time so request_timer_fn() won't
          * kill newly established sessions while we are still trying to thaw