Btrfs: Don't substract too much from the allocation target (avoid wrapping)

When metadata allocation clustering has to fall back to unclustered
allocs because large free areas could not be found, it was sometimes
substracting too much from the total bytes to allocate.  This would
make it wrap below zero.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

fs/btrfs/extent-tree.c

@@ -2255,9 +2255,9 @@ new_group:
 		mutex_unlock(&block_group->alloc_mutex);
 new_group_no_lock:
 		last_wanted = 0;
-		if (!allowed_chunk_alloc && loop > 0) {
-			total_needed -= empty_cluster;
-			empty_cluster = 0;
+		if (!allowed_chunk_alloc) {
+			total_needed -= empty_size;
+			empty_size = 0;
 		}
 		/*
 		 * Here's how this works.
@@ -2277,8 +2277,8 @@ new_group_no_lock:
 			loop++;
 		} else if (loop == 1 && cur == head) {
 
-			total_needed -= empty_cluster;
-			empty_cluster = 0;
+			total_needed -= empty_size;
+			empty_size = 0;
 
 			if (allowed_chunk_alloc && !chunk_alloc_done) {
 				up_read(&space_info->groups_sem);