On 9 April 2015 at 11:21, Paolo Bonzini <pbonz...@redhat.com> wrote: > > > On 09/04/2015 12:14, Peter Maydell wrote: >> On 9 April 2015 at 10:59, Edgar E. Iglesias <edgar.igles...@gmail.com> wrote: >>> > On Tue, Apr 07, 2015 at 09:09:52PM +0100, Peter Maydell wrote: >>>> >> Make address_space_rw take transaction attributes, rather >>>> >> than always using the 'unspecified' attributes. >>> > >>> > Reviewed-by: Edgar E. Iglesias <edgar.igles...@xilinx.com> >>> > >>> > I guess that we eventually will need to convert the dma_ >>> > functions? >> Probably, though I'm not clear what they bring to the party >> that the basic address_space_* functions don't (part of why >> I left them alone). > > At this point, some memory barriers, basically.
So what distinguishes a device that needs the memory barriers and does its accesses via dma_* from a device that doesn't and uses address_space_* or ld/st*_phys ? (Or for that matter a non-device that does memory accesses...) thanks -- PMM