On 30 June 2015 at 21:10, Peter Crosthwaite <peter.crosthwa...@xilinx.com> wrote: > On Tue, Jun 30, 2015 at 12:42 PM, Peter Maydell > <peter.mayd...@linaro.org> wrote: >> The point here is that "do we need to do this" is exactly >> dependent on what we're doing with the CPU. Only if we >> want to put the guest into NS do we do this, and the >> condition for "are we going to put the guest into NS" >> is "is this a Linux boot on a CPU with EL3 but where >> the board says don't boot in S". It matches what the >> existing logic does for when it sets the SCR_NS bit in >> do_cpu_reset() in this file. >> > > Then maybe this belongs on the lowest common denominator for GIC and > CPU - the SoC level. SoCs can register a linux_init function that > checks the CPU and GIC NS support and does the switchup.
The one board we have so far that needs this code doesn't have an SoC level object -- virt just creates the CPU and GIC itself... -- PMM
