On 10/30/2012 09:14 PM, Blue Swirl wrote: > On Tue, Oct 30, 2012 at 11:47 AM, Avi Kivity <a...@redhat.com> wrote: >> Accesses which do not translate will hit the fault region, which >> can then log the access. > > Maybe special casing the fault region should be avoided, the IOMMU > could set up suitable fault regions by itself.
How can it? Please elaborate. I'm not too pleased with using a MemoryRegion for faults, perhaps a MemoryRegionIOMMUOps::fault() callback would be better. -- error compiling committee.c: too many arguments to function