Ocfs2 devel - Jul 2016 - 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")
Signed-off-by: Jun Piao <piaojun at huawei.com>
---
 fs/ocfs2/dlm/dlmcommon.h   |  2 ++
 fs/ocfs2/dlm/dlmmaster.c   | 38 +++-----------------------------------
 fs/ocfs2/dlm/dlmrecovery.c | 30 ++++++++++++++++++++++--------
 fs/ocfs2/dlm/dlmthread.c   | 46 ++++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 73 insertions(+), 43 deletions(-)

diff --git a/fs/ocfs2/dlm/dlmcommon.h b/fs/ocfs2/dlm/dlmcommon.h
index 004f2cb..3e3e9ba8 100644
--- a/fs/ocfs2/dlm/dlmcommon.h
+++ b/fs/ocfs2/dlm/dlmcommon.h
@@ -1004,6 +1004,8 @@ int dlm_finalize_reco_handler(struct o2net_msg *msg, u32
len, void *data,
 int dlm_do_master_requery(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
 			  u8 nodenum, u8 *real_master);
 
+void __dlm_do_purge_lockres(struct dlm_ctxt *dlm,
+		struct dlm_lock_resource *res);
 
 int dlm_dispatch_assert_master(struct dlm_ctxt *dlm,
 			       struct dlm_lock_resource *res,
diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c
index 8c84641..c0b560d 100644
--- a/fs/ocfs2/dlm/dlmmaster.c
+++ b/fs/ocfs2/dlm/dlmmaster.c
@@ -2425,51 +2425,19 @@ int dlm_deref_lockres_done_handler(struct o2net_msg
*msg, u32 len, void *data,
 		mlog(ML_NOTICE, "%s:%.*s: node %u sends deref done "
 			"but it is already derefed!\n", dlm->name,
 			res->lockname.len, res->lockname.name, node);
-		dlm_lockres_put(res);
 		goto done;
 	}
-
-	if (!list_empty(&res->purge)) {
-		mlog(0, "%s: Removing res %.*s from purgelist\n",
-			dlm->name, res->lockname.len, res->lockname.name);
-		list_del_init(&res->purge);
-		dlm_lockres_put(res);
-		dlm->purge_count--;
-	}
-
-	if (!__dlm_lockres_unused(res)) {
-		mlog(ML_ERROR, "%s: res %.*s in use after deref\n",
-			dlm->name, res->lockname.len, res->lockname.name);
-		__dlm_print_one_lock_resource(res);
-		BUG();
-	}
-
-	__dlm_unhash_lockres(dlm, res);
-
-	spin_lock(&dlm->track_lock);
-	if (!list_empty(&res->tracking))
-		list_del_init(&res->tracking);
-	else {
-		mlog(ML_ERROR, "%s: Resource %.*s not on the Tracking list\n",
-		     dlm->name, res->lockname.len, res->lockname.name);
-		__dlm_print_one_lock_resource(res);
-	}
-	spin_unlock(&dlm->track_lock);
-
-	/* lockres is not in the hash now. drop the flag and wake up
-	 * any processes waiting in dlm_get_lock_resource.
-	 */
-	res->state &= ~DLM_LOCK_RES_DROPPING_REF;
+	__dlm_do_purge_lockres(dlm, res);
 	spin_unlock(&res->spinlock);
 	wake_up(&res->wq);
 
-	dlm_lockres_put(res);
-
 	spin_unlock(&dlm->spinlock);
 
 	ret = 0;
 
 done:
+	if (res)
+		dlm_lockres_put(res);
 	dlm_put(dlm);
 	return ret;
 }
diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c
index f6b3138..d926887 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 +2387,18 @@ 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 +2409,17 @@ 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);
diff --git a/fs/ocfs2/dlm/dlmthread.c b/fs/ocfs2/dlm/dlmthread.c
index ce39722..a63e915 100644
--- a/fs/ocfs2/dlm/dlmthread.c
+++ b/fs/ocfs2/dlm/dlmthread.c
@@ -160,6 +160,52 @@ void dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
 	spin_unlock(&dlm->spinlock);
 }
 
+/*
+ * Do the real purge work:
+ *     unhash the lockres, and
+ *     clear flag DLM_LOCK_RES_DROPPING_REF.
+ * It requires dlm and lockres spinlock to be taken.
+ */
+void __dlm_do_purge_lockres(struct dlm_ctxt *dlm,
+		struct dlm_lock_resource *res)
+{
+	assert_spin_locked(&dlm->spinlock);
+	assert_spin_locked(&res->spinlock);
+
+	if (!list_empty(&res->purge)) {
+		mlog(0, "%s: Removing res %.*s from purgelist\n",
+		     dlm->name, res->lockname.len, res->lockname.name);
+		list_del_init(&res->purge);
+		dlm_lockres_put(res);
+		dlm->purge_count--;
+	}
+
+	if (!__dlm_lockres_unused(res)) {
+		mlog(ML_ERROR, "%s: res %.*s in use after deref\n",
+		     dlm->name, res->lockname.len, res->lockname.name);
+		__dlm_print_one_lock_resource(res);
+		BUG();
+	}
+
+	__dlm_unhash_lockres(dlm, res);
+
+	spin_lock(&dlm->track_lock);
+	if (!list_empty(&res->tracking))
+		list_del_init(&res->tracking);
+	else {
+		mlog(ML_ERROR, "%s: Resource %.*s not on the Tracking list\n",
+		     dlm->name, res->lockname.len, res->lockname.name);
+		__dlm_print_one_lock_resource(res);
+	}
+	spin_unlock(&dlm->track_lock);
+
+	/*
+	 * lockres is not in the hash now. drop the flag and wake up
+	 * any processes waiting in dlm_get_lock_resource.
+	 */
+	res->state &= ~DLM_LOCK_RES_DROPPING_REF;
+}
+
 static void dlm_purge_lockres(struct dlm_ctxt *dlm,
 			     struct dlm_lock_resource *res)
 {
-- 
1.8.4.3

On 2016/7/10 18:04, piaojun wrote:
> 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")
> Signed-off-by: Jun Piao <piaojun at huawei.com>
Looks good to me, thanks.

Thanks,
Joseph
> ---
>  fs/ocfs2/dlm/dlmcommon.h   |  2 ++
>  fs/ocfs2/dlm/dlmmaster.c   | 38 +++-----------------------------------
>  fs/ocfs2/dlm/dlmrecovery.c | 30 ++++++++++++++++++++++--------
>  fs/ocfs2/dlm/dlmthread.c   | 46
++++++++++++++++++++++++++++++++++++++++++++++
>  4 files changed, 73 insertions(+), 43 deletions(-)
> 
> diff --git a/fs/ocfs2/dlm/dlmcommon.h b/fs/ocfs2/dlm/dlmcommon.h
> index 004f2cb..3e3e9ba8 100644
> --- a/fs/ocfs2/dlm/dlmcommon.h
> +++ b/fs/ocfs2/dlm/dlmcommon.h
> @@ -1004,6 +1004,8 @@ int dlm_finalize_reco_handler(struct o2net_msg *msg,
u32 len, void *data,
>  int dlm_do_master_requery(struct dlm_ctxt *dlm, struct dlm_lock_resource
*res,
>  			  u8 nodenum, u8 *real_master);
>  
> +void __dlm_do_purge_lockres(struct dlm_ctxt *dlm,
> +		struct dlm_lock_resource *res);
>  
>  int dlm_dispatch_assert_master(struct dlm_ctxt *dlm,
>  			       struct dlm_lock_resource *res,
> diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c
> index 8c84641..c0b560d 100644
> --- a/fs/ocfs2/dlm/dlmmaster.c
> +++ b/fs/ocfs2/dlm/dlmmaster.c
> @@ -2425,51 +2425,19 @@ int dlm_deref_lockres_done_handler(struct o2net_msg
*msg, u32 len, void *data,
>  		mlog(ML_NOTICE, "%s:%.*s: node %u sends deref done "
>  			"but it is already derefed!\n", dlm->name,
>  			res->lockname.len, res->lockname.name, node);
> -		dlm_lockres_put(res);
>  		goto done;
>  	}
> -
> -	if (!list_empty(&res->purge)) {
> -		mlog(0, "%s: Removing res %.*s from purgelist\n",
> -			dlm->name, res->lockname.len, res->lockname.name);
> -		list_del_init(&res->purge);
> -		dlm_lockres_put(res);
> -		dlm->purge_count--;
> -	}
> -
> -	if (!__dlm_lockres_unused(res)) {
> -		mlog(ML_ERROR, "%s: res %.*s in use after deref\n",
> -			dlm->name, res->lockname.len, res->lockname.name);
> -		__dlm_print_one_lock_resource(res);
> -		BUG();
> -	}
> -
> -	__dlm_unhash_lockres(dlm, res);
> -
> -	spin_lock(&dlm->track_lock);
> -	if (!list_empty(&res->tracking))
> -		list_del_init(&res->tracking);
> -	else {
> -		mlog(ML_ERROR, "%s: Resource %.*s not on the Tracking list\n",
> -		     dlm->name, res->lockname.len, res->lockname.name);
> -		__dlm_print_one_lock_resource(res);
> -	}
> -	spin_unlock(&dlm->track_lock);
> -
> -	/* lockres is not in the hash now. drop the flag and wake up
> -	 * any processes waiting in dlm_get_lock_resource.
> -	 */
> -	res->state &= ~DLM_LOCK_RES_DROPPING_REF;
> +	__dlm_do_purge_lockres(dlm, res);
>  	spin_unlock(&res->spinlock);
>  	wake_up(&res->wq);
>  
> -	dlm_lockres_put(res);
> -
>  	spin_unlock(&dlm->spinlock);
>  
>  	ret = 0;
>  
>  done:
> +	if (res)
> +		dlm_lockres_put(res);
>  	dlm_put(dlm);
>  	return ret;
>  }
> diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c
> index f6b3138..d926887 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 +2387,18 @@ 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 +2409,17 @@ 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);
> diff --git a/fs/ocfs2/dlm/dlmthread.c b/fs/ocfs2/dlm/dlmthread.c
> index ce39722..a63e915 100644
> --- a/fs/ocfs2/dlm/dlmthread.c
> +++ b/fs/ocfs2/dlm/dlmthread.c
> @@ -160,6 +160,52 @@ void dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
>  	spin_unlock(&dlm->spinlock);
>  }
>  
> +/*
> + * Do the real purge work:
> + *     unhash the lockres, and
> + *     clear flag DLM_LOCK_RES_DROPPING_REF.
> + * It requires dlm and lockres spinlock to be taken.
> + */
> +void __dlm_do_purge_lockres(struct dlm_ctxt *dlm,
> +		struct dlm_lock_resource *res)
> +{
> +	assert_spin_locked(&dlm->spinlock);
> +	assert_spin_locked(&res->spinlock);
> +
> +	if (!list_empty(&res->purge)) {
> +		mlog(0, "%s: Removing res %.*s from purgelist\n",
> +		     dlm->name, res->lockname.len, res->lockname.name);
> +		list_del_init(&res->purge);
> +		dlm_lockres_put(res);
> +		dlm->purge_count--;
> +	}
> +
> +	if (!__dlm_lockres_unused(res)) {
> +		mlog(ML_ERROR, "%s: res %.*s in use after deref\n",
> +		     dlm->name, res->lockname.len, res->lockname.name);
> +		__dlm_print_one_lock_resource(res);
> +		BUG();
> +	}
> +
> +	__dlm_unhash_lockres(dlm, res);
> +
> +	spin_lock(&dlm->track_lock);
> +	if (!list_empty(&res->tracking))
> +		list_del_init(&res->tracking);
> +	else {
> +		mlog(ML_ERROR, "%s: Resource %.*s not on the Tracking list\n",
> +		     dlm->name, res->lockname.len, res->lockname.name);
> +		__dlm_print_one_lock_resource(res);
> +	}
> +	spin_unlock(&dlm->track_lock);
> +
> +	/*
> +	 * lockres is not in the hash now. drop the flag and wake up
> +	 * any processes waiting in dlm_get_lock_resource.
> +	 */
> +	res->state &= ~DLM_LOCK_RES_DROPPING_REF;
> +}
> +
>  static void dlm_purge_lockres(struct dlm_ctxt *dlm,
>  			     struct dlm_lock_resource *res)
>  {
>

On Sun, 10 Jul 2016 18:04:33 +0800 piaojun <piaojun at huawei.com> wrote:
> 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.
I'm not sure which kernel version this patch is against, but I get
significant rejects when merging into 4.7-rc7.  Can you please redo it
against the latest Linus tree?

Thanks.