On Thu, 2009-05-21 at 14:27 -0400, Becky Bruce wrote:

We have both in every case but one,  
which is in swiotlb_map_page where we call address_needs_mapping()  
without calling range_needs_mapping.

The reason it calls address_needs_mapping without range_needs_mapping is
that in the swiotlb_force=1 case it would trigger every time. If
address_needs_mapping and range_needs_mapping are merged as proposed and
they do not subsume the swiotlb_force check (and I don't think they
would) then I think this will work fine.

It's not actually clear to me that we need that check, though.  Can 
someone explain what case that was designed to catch?

If map_single fails and returns NULL then we try to use
io_tlb_overflow_buffer, if that is somehow not DMA-able (for the
particular device) then the check will trigger. I would have thought we
could arrange that the overflow buffer is always OK and really if
map_page is failing we must be close the edge already.

Ian.