Btrfs: fix an oops when we fail to merge reloc roots

Previously, we will free reloc root memory and then force filesystem
to be readonly. The problem is that there may be another thread commiting
transaction which will try to access freed reloc root during merging reloc
roots process.

To keep consistency snapshots shared space, we should allow snapshot
finished if possible, so here we don't free reloc root memory.
signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Chris Mason <clm@fb.com>

fs/btrfs/relocation.c

@@ -2311,9 +2311,6 @@ void free_reloc_roots(struct list_head *list)
 		reloc_root = list_entry(list->next, struct btrfs_root,
 					root_list);
 		__del_reloc_root(reloc_root);
-		free_extent_buffer(reloc_root->node);
-		free_extent_buffer(reloc_root->commit_root);
-		kfree(reloc_root);
 	}
 }
 
@@ -2355,10 +2352,9 @@ int merge_reloc_roots(struct reloc_control *rc)
 
 			ret = merge_reloc_root(rc, root);
 			if (ret) {
-				__del_reloc_root(reloc_root);
-				free_extent_buffer(reloc_root->node);
-				free_extent_buffer(reloc_root->commit_root);
-				kfree(reloc_root);
+				if (list_empty(&reloc_root->root_list))
+					list_add_tail(&reloc_root->root_list,
+						      &reloc_roots);
 				goto out;
 			}
 		} else {