On Tue, 24 Jan 2023 at 00:02, Richard Henderson <richard.hender...@linaro.org> wrote: > > This is arbitrary, but used by the Huawei TF-A test code.
I guess we should look at how TF-A expects this to work in real-hardware setups. Presumably the idea is that TF-A carves out the RAM it wants to use for the RMM and modifies the device tree / ACPI tables accordingly ? Also, are we going to want to have more link properties for the board to be able to pass in MemoryRegions corresponding to the different physical address spaces, the way we do today for Secure vs NonSecure ? Or do we think we can create suitable board models with "just" a lot of MMUs? thanks -- PMM
