illumos gate

The panic messages in these two range tree functions are accurate if and only if the range tree passed to them is tracking allocated space. In at least the case where we are in space_map_load(…, SM_FREE) via metaslab_load(), the sense is inverted: if we encounter this condition in range_tree_add(), it is because we are freeing a free segment and conversely in range_tree_remove() because we are allocating an allocated one.

We probably should return a status code from these functions that indicates specifically that we failed due to an exact match, and let the caller decide how to communicate the semantics of that failure.