Maybe it's just me, but that seems like just the kind of non-intuitive
behavior that should be documented. :-)
On Mon, Mar 11, 2019 at 2:23 PM Sean Christopherson
<sean.j.christopherson@xxxxxxxxx> wrote:
>
> On Mon, Mar 11, 2019 at 02:19:03PM -0700, Jim Mattson wrote:
> > How does early %rip advancement interact with KVM_GUESTDBG_SINGLESTEP?
> > It seems a bit odd to get the KVM_EXIT_DEBUG for the I/O instruction
> > *before* the KVM_EXIT_IO for the I/O instruction.
>
> It effectively suppresses the single step because vcpu->run->exit_reason
> gets overwritten with KVM_EXIT_IO. kvm_fast_pio() cheeses things and
> calls kvm_fast_pio_{in,out}() even if kvm_skip_emulated_instruction()
> requests an exit.
