This way we don't need to care about the next_index wrapping
around in dma_ops_alloc_addresses.
Signed-off-by: Joerg Roedel <jroedel@suse.de>