> OK, how about this one then?
> > c->busaddr = (__u32) cmd_dma_handle;
> > where cmd_dma_handle is a dma_addr_t (u32 or u64)
It's super-fishy looking but actually I think it's OK, at least as
things stand now. As you see later from how it's freed:
> pci_free_consistent(h->pdev, sizeof(CommandList_struct),
> c, (dma_addr_t) c->busaddr);
this is the bus address of memory from pci_alloc_consistent(), and
since cciss never does pci_set_consistent_dma_mask(), the mask will
remain at the default 32 bits. So the driver is actually safe in
assuming that cmd_dma_handle fits into 32 bits. assuming that
cmd_dma_handle.