On 29/09/2015 05:30, David Gibson wrote: >> Maybe... The only issue is the "granularity" argument, which is >> not in memory_region_register_iommu_notifier. That makes me >> wonder if the replay and registration make sense as separate >> operations. >> >> What about adding a new function memory_region_iommu_replay and >> separate the two phases? > > Hm.. I'm not sure I see much advantage to separating the phases.
It's just that it's (IMO) the clearest API. It shows how the granularity is only used for replay. > But I don't particularly object to the idea either. So, I think it's your > call. Yes, let's separate it. Thanks, Paolo
