--- zzzz-none-000/linux-3.10.107/fs/ocfs2/dlm/dlmrecovery.c	2017-06-27 09:49:32.000000000 +0000
+++ scorpion-7490-727/linux-3.10.107/fs/ocfs2/dlm/dlmrecovery.c	2021-02-04 17:41:59.000000000 +0000
@@ -55,9 +55,6 @@
 static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node);
 
 static int dlm_recovery_thread(void *data);
-void dlm_complete_recovery_thread(struct dlm_ctxt *dlm);
-int dlm_launch_recovery_thread(struct dlm_ctxt *dlm);
-void dlm_kick_recovery_thread(struct dlm_ctxt *dlm);
 static int dlm_do_recovery(struct dlm_ctxt *dlm);
 
 static int dlm_pick_recovery_master(struct dlm_ctxt *dlm);
@@ -208,7 +205,7 @@
 	mlog(0, "starting dlm recovery thread...\n");
 
 	dlm->dlm_reco_thread_task = kthread_run(dlm_recovery_thread, dlm,
-						"dlm_reco_thread");
+			"dlm_reco-%s", dlm->name);
 	if (IS_ERR(dlm->dlm_reco_thread_task)) {
 		mlog_errno(PTR_ERR(dlm->dlm_reco_thread_task));
 		dlm->dlm_reco_thread_task = NULL;
@@ -800,7 +797,8 @@
 				 u8 dead_node)
 {
 	struct dlm_lock_request lr;
-	enum dlm_status ret;
+	int ret;
+	int status;
 
 	mlog(0, "\n");
 
@@ -813,15 +811,16 @@
 	lr.dead_node = dead_node;
 
 	// send message
-	ret = DLM_NOLOCKMGR;
 	ret = o2net_send_message(DLM_LOCK_REQUEST_MSG, dlm->key,
-				 &lr, sizeof(lr), request_from, NULL);
+				 &lr, sizeof(lr), request_from, &status);
 
 	/* negative status is handled by caller */
 	if (ret < 0)
 		mlog(ML_ERROR, "%s: Error %d send LOCK_REQUEST to node %u "
 		     "to recover dead node %u\n", dlm->name, ret,
 		     request_from, dead_node);
+	else
+		ret = status;
 	// return from here, then
 	// sleep until all received or error
 	return ret;
@@ -1071,6 +1070,9 @@
 					     dead_node, dlm->name);
 					list_del_init(&lock->list);
 					dlm_lock_put(lock);
+					/* Can't schedule DLM_UNLOCK_FREE_LOCK
+					 * - do manually */
+					dlm_lock_put(lock);
 					break;
 				}
 			}
@@ -1657,14 +1659,18 @@
 	req.namelen = res->lockname.len;
 	memcpy(req.name, res->lockname.name, res->lockname.len);
 
+resend:
 	ret = o2net_send_message(DLM_MASTER_REQUERY_MSG, dlm->key,
 				 &req, sizeof(req), nodenum, &status);
-	/* XXX: negative status not handled properly here. */
 	if (ret < 0)
 		mlog(ML_ERROR, "Error %d when sending message %u (key "
 		     "0x%x) to node %u\n", ret, DLM_MASTER_REQUERY_MSG,
 		     dlm->key, nodenum);
-	else {
+	else if (status == -ENOMEM) {
+		mlog_errno(status);
+		msleep(50);
+		goto resend;
+	} else {
 		BUG_ON(status < 0);
 		BUG_ON(status > DLM_LOCK_RES_OWNER_UNKNOWN);
 		*real_master = (u8) (status & 0xff);
@@ -1688,6 +1694,7 @@
 	unsigned int hash;
 	int master = DLM_LOCK_RES_OWNER_UNKNOWN;
 	u32 flags = DLM_ASSERT_MASTER_REQUERY;
+	int dispatched = 0;
 
 	if (!dlm_grab(dlm)) {
 		/* since the domain has gone away on this
@@ -1706,17 +1713,28 @@
 			int ret = dlm_dispatch_assert_master(dlm, res,
 							     0, 0, flags);
 			if (ret < 0) {
-				mlog_errno(-ENOMEM);
-				/* retry!? */
-				BUG();
+				mlog_errno(ret);
+				spin_unlock(&res->spinlock);
+				dlm_lockres_put(res);
+				spin_unlock(&dlm->spinlock);
+				dlm_put(dlm);
+				/* sender will take care of this and retry */
+				return ret;
+			} else {
+				dispatched = 1;
+				__dlm_lockres_grab_inflight_worker(dlm, res);
+				spin_unlock(&res->spinlock);
 			}
-		} else /* put.. incase we are not the master */
+		} else {
+			/* put.. incase we are not the master */
+			spin_unlock(&res->spinlock);
 			dlm_lockres_put(res);
-		spin_unlock(&res->spinlock);
+		}
 	}
 	spin_unlock(&dlm->spinlock);
 
-	dlm_put(dlm);
+	if (!dispatched)
+		dlm_put(dlm);
 	return master;
 }
 
@@ -1900,6 +1918,13 @@
 		if (ml->type == LKM_NLMODE)
 			goto skip_lvb;
 
+		/*
+		 * If the lock is in the blocked list it can't have a valid lvb,
+		 * so skip it
+		 */
+		if (ml->list == DLM_BLOCKED_LIST)
+			goto skip_lvb;
+
 		if (!dlm_lvb_is_empty(mres->lvb)) {
 			if (lksb->flags & DLM_LKSB_PUT_LVB) {
 				/* other node was trying to update
@@ -1980,7 +2005,15 @@
 		}
 		if (!bad) {
 			dlm_lock_get(newlock);
-			list_add_tail(&newlock->list, queue);
+			if (mres->flags & DLM_MRES_RECOVERY &&
+					ml->list == DLM_CONVERTING_LIST &&
+					newlock->ml.type >
+					newlock->ml.convert_type) {
+				/* newlock is doing downconvert, add it to the
+				 * head of converting list */
+				list_add(&newlock->list, queue);
+			} else
+				list_add_tail(&newlock->list, queue);
 			mlog(0, "%s:%.*s: added lock for node %u, "
 			     "setting refmap bit\n", dlm->name,
 			     res->lockname.len, res->lockname.name, ml->node);
@@ -1997,11 +2030,8 @@
 	dlm_lockres_drop_inflight_ref(dlm, res);
 	spin_unlock(&res->spinlock);
 
-	if (ret < 0) {
+	if (ret < 0)
 		mlog_errno(ret);
-		if (newlock)
-			dlm_lock_put(newlock);
-	}
 
 	return ret;
 }
@@ -2322,6 +2352,10 @@
 						     dead_node, dlm->name);
 						list_del_init(&lock->list);
 						dlm_lock_put(lock);
+						/* Can't schedule
+						 * DLM_UNLOCK_FREE_LOCK
+						 * - do manually */
+						dlm_lock_put(lock);
 						break;
 					}
 				}
@@ -2346,6 +2380,14 @@
 			} else if (res->owner == dlm->node_num) {
 				dlm_free_dead_locks(dlm, res, dead_node);
 				__dlm_lockres_calc_usage(dlm, res);
+			} else if (res->owner == DLM_LOCK_RES_OWNER_UNKNOWN) {
+				if (test_bit(dead_node, res->refmap)) {
+					mlog(0, "%s:%.*s: dead node %u had a ref, but had "
+						"no locks and had not purged before dying\n",
+						dlm->name, res->lockname.len,
+						res->lockname.name, dead_node);
+					dlm_lockres_clear_refmap_bit(dlm, res, dead_node);
+				}
 			}
 			spin_unlock(&res->spinlock);
 		}
@@ -2710,6 +2752,7 @@
 		     dlm->name, br->node_idx, br->dead_node,
 		     dlm->reco.dead_node, dlm->reco.new_master);
 		spin_unlock(&dlm->spinlock);
+		dlm_put(dlm);
 		return -EAGAIN;
 	}
 	spin_unlock(&dlm->spinlock);