On Mon, Apr 24, 2023 at 02:21:46PM +0300, Viktor Prutyanov wrote:
> +void memory_region_iommu_notify_flags_changed(MemoryRegion *mr)
> +{
> + IOMMUMemoryRegion *iommu_mr;
> +
> + if (mr->alias) {
> + memory_region_iommu_notify_flags_changed(mr->alias);
> + return;
> + }
> + iommu_mr = IOMMU_MEMORY_REGION(mr);
> + memory_region_update_iommu_notify_flags(iommu_mr, NULL);
Do we still want to trap the error if the update failed?
The other question: whether vhost can simply use the existing register /
unregister calls for iommu notifiers, rather than modifying the flags on
its own? I'd assume this happens very rare anyway. Or is there other
concerns?
--
Peter Xu
