On 10/30/2012 03:42 PM, Peter Crosthwaite wrote: > On Tue, Oct 30, 2012 at 7:51 PM, Avi Kivity <a...@redhat.com> wrote: >> On 10/29/2012 10:32 AM, Peter Maydell wrote: >>> On 29 October 2012 06:35, Peter Crosthwaite >>> <peter.crosthwa...@xilinx.com> wrote: >>>> Device model for Primecell PL330 dma controller. >>> >>> A general question -- this is a DMA controller so should it be using >>> the DMAContext APIs now? Avi? >> >> IOMMU functionality will be folded into the memory core, so >> address_space_rw() and family. >> >> > > Hi Avi, > > So does this mean no more dma_contexts and dma_memory_foo?
Correct. > I have > remade the series (and others) using dma_memory_*. Can you point to a > sysbus attached DMA controller currently in tree that does it properly > so we can use it for a basis as there are several (four in total > think) series I have pending all containing DMA which will be effected > by this. I don't think there are any. If your dma controller does not have an IOMMU, all you need to do is replace cpu_physical_memory_*(...) with address_space_*(&address_space_memory, ...). If it does have an IOMMU, then you need to wait until my IOMMU patchset hits mainline (or use DMAContext). -- error compiling committee.c: too many arguments to function