On Thu, Jan 18, 2007 at 01:47:02PM +1100, matthew green wrote:
> 
>    On Thu, Jan 18, 2007 at 05:26:11AM +1100, matthew green wrote:
>    >    
>    >    Thanks for pointing that out.
>    > 
>    > 
>    > good stuff.  i plan to test and commit this in the next couple of days.
>    > i will probably add a __HAVE_BUS_DMATAG_SUBREGION (anyone have a better
>    > name??) define to show this function is available to MI code, and
>    > change bce(4) to barf if this is not present.
>    
>    This probably needs some work for Xen, I can have a look next week-end.
>    This shouldn't be hard to implement.
> 
> what do you think xen will need?  i think all the bus_dma code
> relevant here is shared?

Currenly _xen_bus_dmamem_alloc_range() bails if the memory returned
by UVM isn't in the wanted address range (UVM only knows about pseudo-physical
addresses, so the constraints on DMA address range passed to uvm_pglistalloc()
doens't mean much). There's no way to fix this for Xen2, but we can for Xen3.

-- 
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
     NetBSD: 26 ans d'experience feront toujours la difference
--