btrfs: fix extent_state leak in btrfs_lock_and_flush_ordered_range

btrfs_lock_and_flush_ordered_range() loads given "*cached_state" into
cachedp, which, in general, is NULL. Then, lock_extent_bits() updates
"cachedp", but it never goes backs to the caller. Thus the caller still
see its "cached_state" to be NULL and never free the state allocated
under btrfs_lock_and_flush_ordered_range(). As a result, we will
see massive state leak with e.g. fstests btrfs/005. Fix this bug by
properly handling the pointers.

Fixes: bd80d94e ("btrfs: Always use a cached extent_state in btrfs_lock_and_flush_ordered_range")
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>

fs/btrfs/ordered-data.c

@@ -985,13 +985,14 @@ void btrfs_lock_and_flush_ordered_range(struct extent_io_tree *tree,
 					struct extent_state **cached_state)
 {
 	struct btrfs_ordered_extent *ordered;
-	struct extent_state *cachedp = NULL;
+	struct extent_state *cache = NULL;
+	struct extent_state **cachedp = &cache;
 
 	if (cached_state)
-		cachedp = *cached_state;
+		cachedp = cached_state;
 
 	while (1) {
-		lock_extent_bits(tree, start, end, &cachedp);
+		lock_extent_bits(tree, start, end, cachedp);
 		ordered = btrfs_lookup_ordered_range(inode, start,
 						     end - start + 1);
 		if (!ordered) {
@@ -1001,10 +1002,10 @@ void btrfs_lock_and_flush_ordered_range(struct extent_io_tree *tree,
 			 * aren't exposing it outside of this function
 			 */
 			if (!cached_state)
-				refcount_dec(&cachedp->refs);
+				refcount_dec(&cache->refs);
 			break;
 		}
-		unlock_extent_cached(tree, start, end, &cachedp);
+		unlock_extent_cached(tree, start, end, cachedp);
 		btrfs_start_ordered_extent(&inode->vfs_inode, ordered, 1);
 		btrfs_put_ordered_extent(ordered);
 	}