Hi, > > > (b) pass MachineState instead of ConfidentialGuestSupport, so > > > we can use the MachineState here to generate the madt. > > > > > >Luigi, any opinion? I think device tree support will need access to > > >MachineState too, and I think both madt and dt should take the same > > >approach here. > > > > I have a slight preference over MachineState as it's more generic and we > > don't need to add more fields in IgvmCfg for new features. > > > Passing in MachineState would be easy, but do we really want to add machine > related logic (building of ACPI tables, and later maybe device trees) > into the igvm backend?
Good point. That clearly speaks for (b). There already is MachineState->fdt, filled in by machine code. We can let machine code store the madt in MachineState too, and the have the igvm code simply pick it up from there. take care, Gerd
