On 15/01/20 23:51, Oliver Upton wrote: > Good point. I'm instead inclined to call the hook to emulation_complete > (will rename as appropriate) from kvm_vcpu_do_singlestep(), as it > appears this is how x86_emulate_instruction processes the trap flag. > Need to take a deeper look + test to ensure this change will fix MTF for > full instruction emulation. Cool, I was going to suggest the same. You can use the forced emulation prefix to send your testcase down the x86_emulate_instruction path. Paolo
