On Tue, Jan 10, 2023 at 01:57:05PM +0100, Borislav Petkov wrote:
I'm doubtful now as you're injecting errors so you're not really in #MC contextI did some more thinking:
but in this contrived context which is actually an NMI one. So we need to think
about how to fix this case.
*if* this really is a real issue - and not some silly qemu games - then
native_machine_crash_shutdown() does all the cleanup before the kdump kernel is
started.
Any NMI clearing, maybe using iret_to_self() etc, #MC resetting etc should
happen there and not anywhere else.