drbd: fix "stalled" empty resync

With sync-after dependencies, given "lucky" timing of pause/unpause
events, and the end of an empty (0 bits set) resync was sometimes not
detected on the SyncTarget, leading to a "stalled" SyncSource state.

Fixed this by expecting not only "Inconsistent -> UpToDate" but also
"Consistent -> UpToDate" transitions for the peer disk state
to end a resync.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>

2 files changed, 14 insertions, 5 deletions

diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c
index e64b1c897c82..16c3710e1b9c 100644
--- a/drivers/block/drbd/drbd_nl.c
+++ b/drivers/block/drbd/drbd_nl.c
@@ -2262,7 +2262,7 @@ int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info)
 
         if (rs.no_resync && mdev->tconn->agreed_pro_version < 93) {
                 retcode = ERR_NEED_APV_93;
-                goto fail;
+                goto fail_ldev;
         }
 
         rcu_read_lock();
@@ -2272,7 +2272,7 @@ int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info)
                 new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL);
                 if (!new_disk_conf) {
                         retcode = ERR_NOMEM;
-                        goto fail;
+                        goto fail_ldev;
                 }
         }
 
@@ -2310,6 +2310,10 @@ int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info)
  fail:
         drbd_adm_finish(info, retcode);
         return 0;
+
+ fail_ldev:
+        put_ldev(mdev);
+        goto fail;
 }
 
 int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info)
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 4004a682d0ec..aba04d7dadf5 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -3748,9 +3748,14 @@ static int receive_state(struct drbd_tconn *tconn, struct packet_info *pi)
         os = ns = drbd_read_state(mdev);
         spin_unlock_irq(&mdev->tconn->req_lock);
 
-        /* peer says his disk is uptodate, while we think it is inconsistent,
-         * and this happens while we think we have a sync going on. */
-        if (os.pdsk == D_INCONSISTENT && real_peer_disk == D_UP_TO_DATE &&
+        /* If this is the "end of sync" confirmation, usually the peer disk
+         * transitions from D_INCONSISTENT to D_UP_TO_DATE. For empty (0 bits
+         * set) resync started in PausedSyncT, or if the timing of pause-/
+         * unpause-sync events has been "just right", the peer disk may
+         * transition from D_CONSISTENT to D_UP_TO_DATE as well.
+         */
+        if ((os.pdsk == D_INCONSISTENT || os.pdsk == D_CONSISTENT) &&
+            real_peer_disk == D_UP_TO_DATE &&
             os.conn > C_CONNECTED && os.disk == D_UP_TO_DATE) {
                 /* If we are (becoming) SyncSource, but peer is still in sync
                  * preparation, ignore its uptodate-ness to avoid flapping, it