On Sat, Dec 15, 2018 at 12:29:11AM +1100, Michael Ellerman wrote: > I think the problem is that we don't want to set iommu_bypass_supported > unless cell_iommu_fixed_mapping_init() succeeds. > > Yep. This makes it work for me on cell on top of your v5. Thanks, this looks good. I've folded it with the slight change of moving the iommu_bypass_supported setup into cell_iommu_fixed_mapping_init.
