ocfs2/dlm: continue to purge recovery lockres when recovery master goes down

We found a dlm-blocked situation caused by continuous breakdown of
recovery masters described below.  To solve this problem, we should
purge recovery lock once detecting recovery master goes down.

N3                      N2                   N1(reco master)
                        go down
                                             pick up recovery lock and
                                             begin recoverying for N2

                                             go down

pick up recovery
lock failed, then
purge it:
dlm_purge_lockres
  ->DROPPING_REF is set

send deref to N1 failed,
recovery lock is not purged

find N1 go down, begin
recoverying for N1, but
blocked in dlm_do_recovery
as DROPPING_REF is set:
dlm_do_recovery
  ->dlm_pick_recovery_master
    ->dlmlock
      ->dlm_get_lock_resource
        ->__dlm_wait_on_lockres_flags(tmpres,
	  	DLM_LOCK_RES_DROPPING_REF);

Fixes: 8c0343968163 ("ocfs2/dlm: clear DROPPING_REF flag when the master goes down")
Link: http://lkml.kernel.org/r/578453AF.8030404@huawei.com
Signed-off-by: Jun Piao <piaojun@huawei.com>
Reviewed-by: Joseph Qi <joseph.qi@huawei.com>
Reviewed-by: Jiufei Xue <xuejiufei@huawei.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 21 insertions, 8 deletions

diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c
index f6b3138..dd5cb8b 100644
--- a/fs/ocfs2/dlm/dlmrecovery.c
+++ b/fs/ocfs2/dlm/dlmrecovery.c
@@ -2343,6 +2343,7 @@ static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node)
 	struct dlm_lock_resource *res;
 	int i;
 	struct hlist_head *bucket;
+	struct hlist_node *tmp;
 	struct dlm_lock *lock;
 
 
@@ -2365,7 +2366,7 @@ static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node)
 	 */
 	for (i = 0; i < DLM_HASH_BUCKETS; i++) {
 		bucket = dlm_lockres_hash(dlm, i);
-		hlist_for_each_entry(res, bucket, hash_node) {
+		hlist_for_each_entry_safe(res, tmp, bucket, hash_node) {
  			/* always prune any $RECOVERY entries for dead nodes,
  			 * otherwise hangs can occur during later recovery */
 			if (dlm_is_recovery_lock(res->lockname.name,
@@ -2386,8 +2387,17 @@ static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node)
 						break;
 					}
 				}
-				dlm_lockres_clear_refmap_bit(dlm, res,
-						dead_node);
+
+				if ((res->owner == dead_node) &&
+							(res->state & DLM_LOCK_RES_DROPPING_REF)) {
+					dlm_lockres_get(res);
+					__dlm_do_purge_lockres(dlm, res);
+					spin_unlock(&res->spinlock);
+					wake_up(&res->wq);
+					dlm_lockres_put(res);
+					continue;
+				} else if (res->owner == dlm->node_num)
+					dlm_lockres_clear_refmap_bit(dlm, res, dead_node);
 				spin_unlock(&res->spinlock);
 				continue;
 			}
@@ -2398,14 +2408,17 @@ static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node)
 				if (res->state & DLM_LOCK_RES_DROPPING_REF) {
 					mlog(0, "%s:%.*s: owned by "
 						"dead node %u, this node was "
-						"dropping its ref when it died. "
-						"continue, dropping the flag.\n",
+						"dropping its ref when master died. "
+						"continue, purging the lockres.\n",
 						dlm->name, res->lockname.len,
 						res->lockname.name, dead_node);
+					dlm_lockres_get(res);
+					__dlm_do_purge_lockres(dlm, res);
+					spin_unlock(&res->spinlock);
+					wake_up(&res->wq);
+					dlm_lockres_put(res);
+					continue;
 				}
-				res->state &= ~DLM_LOCK_RES_DROPPING_REF;
-				dlm_move_lockres_to_recovery_list(dlm,
-						res);
 			} else if (res->owner == dlm->node_num) {
 				dlm_free_dead_locks(dlm, res, dead_node);
 				__dlm_lockres_calc_usage(dlm, res);