I thought we had an agreement for a long time already that the use of
dma64_addr_t in the sym53c8xx_2 driver was wrong, but it's still
there.